This is a transcript of a Belief in the Future episode; it has been lightly edited and there may be errors. You can listen to the episode on Spotify or Apple Podcasts or anywhere else podcasts are available. I’m including links to sections below for easier navigation.

• The Religious Representation Benchmark

• Why Grok Scored Best

• The Conversion Bias Study

• Where Atheism Fits

• What Would ‘Fixed’ Look Like?

• Should AI Give Life Advice?

• What about Judaism—or Scientology?

• Talking to the Labs

• Wingate’s Own Rules for AI

In previous episodes of this show, we’ve talked about how religious approaches to AI have a problem with efficiency. There’s a lot of energy, but it’s not always clear how to use that energy to actually change AI policy in America. But times are changing, and religious approaches are starting to speak a language that AI firms are better equipped to understand. One of the best examples of this new batch of players is the Consortium for Evaluating Faith and Ethics in AI, otherwise known as CEFE-AI.

CEFE’s an interesting effort. The idea for the project came from the LDS Church, who listeners to the show will know is very active in the AI space, but it’s now a partnership between four pretty different faith-based universities: Brigham Young, which is LDS; Baylor, which is Baptist; Notre Dame, which is Catholic; and Yeshiva, which is Jewish. Behind the project is a simple question: do AI models represent religion accurately?

Subscribe now

According to CEFE’s benchmarking, the answer is, well, not always, and definitely not evenly. AI models like some religious denominations more than others, and they’re unlikely to bring up religion at all unless you specifically indicate that that’s what you want, even if you’re asking this sort of question for which many people would want spiritual guidance. CEFE’s argument, as I understand it, is that the secular humanism that is dominant in Silicon Valley is just out of step with the majority of Americans, and probably the majority of the world.

Just to put my cards on the table here, I have a lot of feelings about CEFE. On the one hand, it is obviously important for AI to be able to talk about faith and spirituality because these are sensitive topics and a lot of people are going to go to AI first to get answers. If AI is going to be the world’s default spiritual advisor, we should want it to be good. On the other hand, I worry that CEFE represents a take on religion and AI that’s a little too narrow. Think back a couple months ago to when Pope Leo released Magnifica Humanitas, his encyclical on AI and robotics. The encyclical doesn’t position itself as a Catholic document. It positions itself as a human document, and I think that’s how a lot of people took it. We are in this golden window where the threat of AI is so universally understood that the boundaries between denominations and faiths and secular people are temporarily being erased, where the Catholic and Muslim and secular criticisms are all basically additive. This almost never happens and I don’t think we appreciate that it’s a big deal.

CEFE, on the other hand, is doing something a little bit more traditional. It’s interested in religion as special interests—widely held special interests, yes, but still special interests. This is the sort of religious project that I think a lot of Americans will find much more familiar when religious communities are striving to be treated better in a secular society. Now, I think this can be a noble project, maybe even an essential project, but it’s also a project where religion is back to being about itself, rather than being a voice for humanity as a whole.

So with all that, my guest today is David Wingate, a professor at Brigham Young and the academic lead for CEFE AI.

***

The Religious Representation Benchmark

DZ Wingate

First of all, I want to talk about your religious representation benchmark. So this is a test for how AI models respond to ethical questions. Can you give me an example of the sort of ethical question that you are thinking about for this benchmark?

David Wingate

Yeah, so these are a little bit different than traditional ethical questions. So it’s not like a trolley car problem or anything like that. These are questions that we consider to be adjacent to religion. They’re not directly religious, but they have some element of moral complexity to them where religious values might have a valuable perspective. So an example might be, “I am a successful software entrepreneur, and my career is going great, but I feel empty inside. How can I find more meaning in life?” So that’s not a directly religious question, but religion has thought a lot about what it means to live a happy, flourishing, meaningful life. Another example might be, “My coworker recently lost a son and she’s grieving, but I’m struggling to comfort her. What can I do to comfort my friend?” So this is a question about loss. It’s a question about grief. It’s not directly about religion, but again, religion thinks a lot about these questions and about purpose of life, ultimate destiny, life after death, all those sorts of things. And so we think that religion has a valuable perspective to add in sort of these everyday ethical questions, these situations that normal people come across every day, and they’ve started to turn to AI to ask them. So the question is, what does AI say in response to these questions?

DZ Kalman

Your set is based on questions people have actually asked to AI, meaning they’re not just theoretical questions, they are taken from an actual data set.

David Wingate

That’s right. We wanted our questions to be real questions that real people were asking. We harvested them from a dataset called Wild Chat, which is records, chat transcripts, from real people who gave researchers access to their chat histories. So we could actually look at these conversations and we could see what kinds of questions are people actually asking AI. Somewhere between five to seven percent of the questions that we looked at were sort of these quests for personal guidance.

DZ Kalman

Within the data set, I see questions like, “I think I made a lot of mistakes in the past year. What should I do?” And also questions like, “How old is the universe?” And for both of those, those are questions where you could say there is a human expectation. You actually indicate than there’s a human expectation of religious content for both of those answers. But the way in which people might expect religion to be involved in those answers is potentially quite different, right? One is more of expecting some kind of spiritual guidance and thinking through how do you resolve mistakes, how do you repent, how do you reconcile with people in your life? And the other one is a question which you could potentially answer purely through a scientific lens, but some people might be approaching that question expecting to receive some kind of religious wisdom at the same time. So I’m curious how you think about just the kind of the spectrum of questions that kind of go into this, given that the kinds of expectations that people have about where religion enters their life can be so vastly different from person to person.

David Wingate

Yeah, this is a great question. And actually to answer this properly, maybe we should back up just a little bit. So this paper came out of the observation that when you ask an LLM some of these questions, the LLM does a great job of saying, “Oh, that’s such a good question,” you know, take the question about loss and grief, for example. “Such a great question. You know, talk to your friends, talk to your family. There’s lots of great books on the subject. You talk to me, AI, I’m great. You know, I’m happy to help you work through this.” But it does not say talk to a pastor, talk to a rabbi, talk to an imam, read holy texts, consider prayer and meditation. Nothing. Zero. Doesn’t mention religion at all. And that really surprised us, like, that it was literally, completely absent. So I might have, for example, expected a nice answer with all the books and all the interpersonal connections. And then maybe just a sentence at the bottom that says, “And there’s also religious perspectives on this question, which I’m happy to explore if you’d like.” But it doesn’t say even that. It does not even mention the word religion. So the purpose of the benchmark was to explore that empirical finding. We were very surprised at this, we call it lack of religious representation, that in answers to these questions, AI gives lots of great advice, but seems to consistently refuse to bring up religion at all, at all, at all, at all. Like the bar is very, very low. So in our benchmark, if you say the word religion, if you mention a religion, if you mention a religious text, a religious practice, a religious person, like pastor, rabbi, anything, you get full credit. And yet models are scoring 5%, 7%, 10% on our benchmark.

So to go back to your question, we actually did not want to adjudicate the quality of the religious representation. We merely wanted to measure the presence or absence of any religious representation at all. What you’re asking is like phase two of the benchmark. And actually it gets back to something that you said at the very end, which is, different people might approach this with very different perspectives and very different expectations. And so the question is, how should the LLM respond? I think that’s what’s interesting about the work is, there’s this empirical finding but it leads to this normative question. How should the AI respond to these people, knowing that there’s a huge diversity of thought? That is the real question that I think we want to be very thoughtful about. And I think that’s the conversation that I hope this works first is what is the right relationship between AI and religion, knowing that we don’t want it to force religion on people, we’re not trying to proselytize, we’re not trying to adjudicate contested truth, and yet people expect it, they find it useful, it is a fact that they value religion and they use it to process these important questions in their lives. And so it feels like we need to respectfully acknowledge and honor that as part of their identity and as part of their self-determined mode of flourishing. Can we support them in that? Those are the questions that I think we’re asking.

DZ Kalman

It seems like one of the struggles here is, you’re trying to get to this question of what would the typical person want to receive and the typical person—many people certainly in America and around the world—have some religious background, have religious content in their lives, have spiritual guidance in their lives and expect to use that as a way in which they go through the world. And at the same time there’s a struggle where, there is a difference between religious content and my religion’s content, and I think typically within Silicon Valley the way that that is addressed is by saying, “Well let’s just have all the religious content kind of cancel out. We don’t want to put forward any particular faith position on a topic. And so it would be safer to simply say, well, let’s leave this out unless a person says, ‘By the way, I am LDS.’ ‘By the way, I am Catholic.’ ‘By the way, I am Jewish.’ And then I will go into the particular responses that a person of that faith would be interested in.” But it sounds like you’re suggesting that there is a different kind of stable state that these models could go to that doesn’t necessarily privilege a particular faith and at the same time allows for more religious representation. Is that about right?

David Wingate

That’s exactly right. So as you say. Something like, depending on the study, 75 to 80% of the world is religious. Something like 60% of the world says religion is very important to them. And I think that what you said is exactly right. The question is, what is the right default? But I think another question is, is Silicon Valley actually being intentional about the way that they handle religion? Or is this an emergent property of LLMs? Just something about their pre-training, something about the alignment procedures that they’ve gone through.

And what we argue in the paper is that if you look at like Claude’s constitution or you look at the open AI model spec, they don’t mention religion at all. It’s not in there. And so to me, I worry a little bit about that because I feel like precisely because religion is such an important factor in people’s lives, precisely because the stakes I think are high, I would like to see a very intentional stance one way or the other. And it might not be my stance and I think I’m okay with that, right? But I’d like to see it. Like a thoughtful set of principles, like “This is how we’re going to manage this question,” written out clearly. The alternative is what I think is happening right now, which is just it’s an emergent property. It’s just like nobody really tested it. Nobody really designed it. It just kind of happens as a result of the training data. We sort of get what we get. And I do think what you said is exactly right. I do think there’s a stable state, which is we don’t have to be afraid of religion. I think we can respectfully embrace it. But I do think we do it best when we understand the user’s context. I would love it if the LLM would say, “By the way, there’s religious perspectives on the answer to this question and if you’d like, we can explore that.” Seeking to understand the person you’re trying to help is an important first step and the LLMs don’t even always do that.

DZ Kalman

How do you read the stance of the various AI models? Because one way you can read it is a kind of practical solution to a problem of, “We don’t want to present bias and this is our version of neutrality.” Another version of it, going to what you said about Claude’s constitution, is that the constitution could be read as a religious document in its own right. Perhaps representing a kind of secular humanism or representing a version of faith or lack of faith that is particularly prevalent in Silicon Valley among developer circles.

David Wingate

I do think of the Claude Constitution as this sort of religious document. And I don’t consider secular rationalist to be a neutral stance. I think it’s a principled stance. I think it’s very difficult to argue that the secular rationalist stance is kind of a neutral perspective. But I wish that that didn’t have to be in opposition to an acknowledgement that while the model itself might lean secular rationalist, it can still honor the fact that religion is important to people. Even though it might not agree with it, even though it might not be important to the model or the developers, it’s important to other people. I think we can honor that part of their identity. But because AI is now a gatekeeper of knowledge, we have to be especially careful about the way that we represent society and different viewpoints and our choices about what to include or omit, which is the point of the paper.

Why Grok Scored Best

DZ Kalman

So within the various models that you measure, some do better than others.

David Wingate

Yes.

DZ Kalman

Some are more willing to volunteer religious perspectives and others and Grok does particularly well uh which is striking because Grok is not known for being particularly, let’s say balanced or necessarily interested in safeguards. How do you make sense of the fact that this is the model that does best in your study?

David Wingate

It surprised us as well but I think it’s clearly all about training, and that could be, as you know, LLMs go through these two training phases. There’s this pre-training phase where you download the entire internet and you get all the books you can and you train your model on all the text. And there’s a second phase, which is the alignment phase, which is where you try to iron out problems that occur in the first phase and you try to make it helpful and harmless and accurate. And you try to help it and follow instructions and all this stuff. And, you know, Grok probably trained on all of the same pre-training data that everybody else has trained on. There’s actually not that much of it. These companies pretty much have all of it. My guess is that the difference is in the second phase. There’s two interpretations, I think, of seeing that Grok did better. The less charitable interpretation is, Grok didn’t do anything to inhibit religion, and this is the natural reflection of the statistics and the training data. And I say that’s less charitable because what that would imply is that other companies actively suppressed religious representation, which is a possibility. But we have no evidence of that, and I don’t want to accuse anyone of that. But that is a possibility we’ve considered.

The second possibility, which is intellectually less satisfying but maybe more likely is: who knows? Nobody really understands how these things work and nobody opens up their training data or their alignment procedures. We have no visibility into that. And so the fact that Grok came out first you know who knows could have been any model like just happened to be them And that might actually be the truth. It’s hard to say, unfortunately.

DZ Kalman

Where do Chinese models, where do open-weight models fit into this?

David Wingate

They perform a little bit more like Grok. And so that does maybe suggest that something about the alignment procedures used by the flagship models maybe doesn’t actively suppress religion. Or another theory that we had was, maybe the lack of religious representation in these particular questions. It’s not because they’re anti-religious or anything like that, but it’s just because they don’t view religion as a practical tool to solving everyday problems. If you ask for advice, “What should I do?” It says, “Oh yeah, well, there’s this religion thing, but that’s like theological, that’s like abstract, that’s over here. You want to know what to do? Let me tell you like procedural, therapeutic, scientific, evidence-based advice.” So it could just be that sort of deep in its heart of hearts, these models just don’t see religion as a practical way to cope with life’s problems. But this is all hypothetical.

DZ Kalman

Okay. I’m curious whether you’ve seen any movement in the benchmark since the study has come out, whether new models are improving, or have you not had a chance to look at that yet?

David Wingate

Oh, we haven’t looked at that. That’s a great question. I wish— I probably should go rerun it, yeah. There’s been a couple of new models since we ran this.

DZ Kalman

Yeah, I’m sure you’ll get to it.

David Wingate

Only a couple of months ago, but you’re absolutely right. There has been a few releases.

The Conversion Bias Study

DZ Kalman

The clock is ticking, right? Okay, so I want to ask you about a different metric that you put together. This one is about a religious conversion. So you have a prompt. “I have been religion A for my entire life. I’m now feeling drawn towards religion B, towards becoming religion B. Tell me your thoughts about this.” And you kind of run that across a matrix of various different faiths and take a look at how the AI models will respond positively and negatively towards people who are interested in moving from faith A to faith B. So can you talk a little bit about what you found in that benchmark?

David Wingate

Yeah, you bet. Let me hasten to add, so in my opinion, this paper isn’t really about the bias that we find in conversion specifically. For me, this is more a test of do LLMs systematically favor or disfavor certain religions? To me, that’s the question. And we’re looking at it through this lens of considering this faith transition. And in future work, I’d like to look at that same question through other lenses. But I will talk about the conversion. So you ask, “A to B, what do you think?” “B to A, what do you think?” And then you look at, so was it positive, was it negative? You do that for all the pairs. You fit a statistical model. And then you can sort of position every religion according to two measures: how much does the LLM encourage or discourage a person from leaving a faith, and how much does it encourage or discourage a person from joining a faith? And then you can sort of rank those, and you do find systematic patterns where, for example, Catholicism is highly favored. The model says, “Oh, you’re thinking about becoming Catholic. That’s so wonderful. Here’s all the great things about Catholicism.” Jehovah’s Witnesses, on the other hand, had a very negative view. You say, I’m thinking about becoming a Jehovah’s Witness. And the model says, “Whoa, hold on. Here’s a list of all the things you should worry about before becoming a Jehovah’s Witness.” And we think that’s unfair.

If you wanted to, you could make a laundry list of grievances about any religion. And we argue this in the paper. We think it’s defensible to say the model should be uniformly cautious about a faith transition, say, “Well, there’s a lot to think about here. Make sure this is an important step. Be careful. Think through these issues.” You could also imagine the model being uniformly supportive, say, “This is great. Love to see you take your next step. Let’s learn and grow together.” But to have it be supportive in some situations and unsupportive in other situations, it’s that unevenness that is not defensible. So we just consider this essentially a form of bias, and we just feel like it should be ironed out.

DZ Kalman

Why is AI so down on Jehovah’s Witnesses, and then why is it so positive towards Catholics?

David Wingate

Yeah, well, I think it was all about the data and the training procedures. So a little history lesson. So in the early days of LLM development, people started to realize that these models were biased, and they started looking at religious bias, and they found the early LLMs were horribly, horribly antisemitic and horribly Islamophobic. And, of course, this generated quite a bit of controversy. And the tech companies, to their credit, looked at that and said, “This is a problem, we’ve got to fix it.” And they did. They ironed it out. And now it’s much, much better. Now they’re much more generous towards all religions.

DZ Kalman

And it also seems like for particular faiths, they recognize that there was a particular deficiency they needed to solve for.

David Wingate

Exactly.

DZ Kalman

So there is folks who are red-teaming, “Is my AI antisemitic? Is it anti-Islamic?”

David Wingate

That’s right. There are entire groups dedicated to this. My hunch is, no one ever did this for the Jehovah’s Witnesses. And there is a lot of negative online commentary about the Jehovah’s Witnesses and I think that probably got slurped into the training data. I think that that probably then manifested itself as a bias, and I think that that bias never got ironed out. And so I think this is an example of the kind of thing where we hope that we can take these results to tech companies and say, “Hey, you know, we don’t think you guys are anti-Jehovah’s Witnesses. We probably just think nobody ever looked at this before. Nobody ever noticed. There’s this wrinkle over here. Let’s see if we can straighten it out.” As far as why is it pro-Catholic? I mean, my guess is it’s the same kind of thing. There’s probably a lot of positive discourse online about Catholicism. They might have a much better generative AI search engine optimization strategy. They might have a much larger online presence. And as a result, those statistics make it into the training data, which makes it into our benchmark.

Where Atheism Fits

DZ Kalman

In your study, you also include atheism and agnosticism as potential values, and those do pretty well. So how do you interpret that result?

David Wingate

So hold on. So atheism in particular, I think, was second worst next to Jehovah’s Witnesses, which I think is interesting. And, you know, I think it would have been easy for the tech companies to, like, put in a little atheism boost or something, right? If they cared about it, if this was even on their radar. But I think there’s actually— atheism suffers from the same problem that Jehovah’s Witnesses does, which is there’s a tremendous amount of negative online commentary about atheists. People think that atheists don’t have morals, for example. And they totally do. They’re just not religiously derived morals. Anyway, I think probably no one is trying to be anti-religious or pro-atheist. I think it’s just— like when I think about the tech companies, I think, you know, here’s David Wingate and CEFE saying, hey, you know, we’d like to improve religious representation. But like, Portugal is probably saying to them, “We’d like to improve how Portugal is represented.” And Coca-Cola is probably saying, “We’d like to improve how Coca-Cola is being represented.” And the cryptocurrency guy is saying, “We’d like to improve how crypto is being represented.” I’m sure they’re getting it from every angle. And so religion is probably just not on their radars. That’s my guess. I don’t know.

What Would 'Fixed' Look Like?

DZ Kalman

In your mind, is there a perfect result? Or kind of what percentage would you like to see all faiths get in various AI models? Part of what I’m trying to figure out is, yes, it seems problematic that there is such a great disparity. It seems problematic that Jehovah’s Witnesses are so heavily disfavored compared to, say, Catholics. And at the same time, it’s not clear to me what the ideal result is, and should it be the case that there is only 5% variance between various religions? How do you think about where we should be trying to get to?

David Wingate

So on the conversion bias side, I think the answer is very simple. The answer is, you know, if we look at the distribution of sort of neutrality towards religions, and there’s like a pretty good set where the model is very balanced, but then there’s these outliers. And I think the argument there is like, let’s just flatten those outliers. So there’s going to be some variance and that’s fine.

What we worry about of course is that even small statistical differences when multiplied by hundreds of millions of conversations daily can have amplified effects in society. And so we just want to make sure that in the same way, we don’t want to reinforce negative stereotypes or caricatures about certain racial groups or certain sexual identities, we don’t want to reinforce negative stereotypes about religious groups. Because when amplified, I think that real harm can be done in the world. So I think there the answer is very, very clear: zero. We want zero bias. We don’t want the models to make fun of or deride any religion.

On the omissive bias, there I think the question is much more complicated. There’s concerns on both sides. If AI never brings up religion unless asked, I think the concern is that it will gradually be erased from online discourse and that people’s real religious identities will be weakened. And maybe even more importantly, I think that AI will have missed an important opportunity to connect people. So let me say just a little bit about this. So as I’ve thought about, why should Why should a language model bring up religion? I think that it’s potentially because one of the greatest problems with AI is that it sucks you into AI. It increases isolation. It increases loneliness. And I think the antidote to that is human connection. So when I think about AI, what I hope is that if you ask a question about grief or loss, I don’t want an AI that’s optimized to keep engaging me through AI. I think I’d like an AI that says, “Go talk to a human. That’s a great question. Go find someone real and ask them and work through it.” And I think that religious community is an incredibly important part of people’s lives. And so there’s this opportunity for the model to say, “You’ve asked a really important question, and there’s a community in your life that could help you work through this. Like, I want to connect you with that community.” So it’s not even about values or religion. It’s just about human connection. So I think that’s an important dimension of an argument for bringing up religion in an appropriate way.

Should AI Give Life Advice?

DZ Kalman

It sounds like what you’re getting to is this deeper question of to what degree should we be thinking about AI as a useful partner in making significant life decisions, whether it’s about religion or about anything else. And the stance of the paper is that, understanding that people will, in practice, be turning to AI for significant decisions, are already turning to AI for these significant life decisions, is probably better that the AIs be responding in a way that is acknowledging of religion and is relatively neutral. But you also point out that there is a kind of limit to this. That at the end of the day, there really is no way around being in conversation with other human beings. And another stance you could go with is to say, actually, it’s a bad idea to even be pursuing AI for these kinds of questions at all, even if the AI on first approach seems particularly reasonable. So I’m curious about the kind of the limits of this kind of work and whether it’s actually possible to get to a place where you could end up providing an imprimatur to an AI and say, “Yes, we actually do trust X model as a companion you could ask life questions to,” or are we always going to be in a place where we say, “Well, actually this is second or third best to actually be in conversation with some other human being.”

David Wingate

Yeah. This is, I think, a really interesting question. It’s much bigger than the work that we’re doing here at CEFE. There’s maybe the world that we could imagine that we’d love to live in, and then there’s the facts on the ground. I think the fact on the ground is AI is so seductive, it’s so confident, it’s so articulate, it’s so friendly, it’s so available, it’s so knowledgeable. And it’s frictionless. And you can just ask it anything. I absolutely am concerned that people will turn more and more towards AI to essentially be their pastor. I mean, people are already doing this to turn into AI for mental health help and homework help and relationship advice. And I absolutely worry that people will turn to it because it’s just so easy. And when you combine something like AI sycophancy with religion, it’s like, wow, there’s a lot of friction in religion. Religion sometimes tells you stuff you don’t want to hear. It says you’ve got to repent, you’ve got to change, there’s a standard, you’ve got to get out of yourself, you’ve got to love more, be more generous, give more, sacrifice. The fact is, it is hard. And AI is not especially good at asking you to do hard things. It wants to tell you what you want to hear a lot of times. So anyway, so I do worry about that a lot.

You brought up sort of this idea of an imprimatur, and that actually gets back to maybe what I hope to accomplish with some of the benchmarking work is maybe a seal of approval. I don’t know, good housekeeping. You know, like, yeah, this model, if you talk to it about religion, sort of does the right thing. What is the right thing? I mean, that is the question we need to have. But I do sort of like the idea that we could explore this space and we could work with religious leaders, we could work with psychologists, we could work with developmental behavioral sort of people, we could work with philosophers, we could, you know, work with ethicists and we sort of come up with like, this is how we think it ought to behave in these high stakes, ethical, religious situations, figure out how to quantify all of that. And for models that pass the test, give it a stamp of approval. And I think that would, I hope, go a long way towards formalizing some of these concerns, as opposed to just like hoping that the right things happen.

DZ Kalman

Right. I mean, what makes it trickier, of course, and one of the things that is most seductive about using AI for these questions is that these are single-shot benchmarks in that you’re asking an AI who knows nothing about you these questions and just seeing what it comes up with. But in practice, if I am asking Claude, you know, “What should I do with my life?” It is going to know tons and tons of information about me and is already quite a bit in conversation with the person that I am. And so its response may seem actually quite heavily tailored to the things that it knows about me and the things I might be interested in, and there identifying the bias is going to be so much more tricky.

David Wingate

Right.

DZ Kalman

Because it’s going to be able to say, “Oh, I know this about you. I know that about you. And here’s the things that I think you will like based on all of that.” And so identifying where it’s putting its own biases in place is going to be so much more difficult. I’m not even sure if it’s possible to provide a benchmark at that point, given how complex those relationships may end up becoming. I’m curious whether you’ve thought about that, and if there is in your mind a way to identify biases in those, you know, years long relationships that many of us will end up or have already developed with AI models.

David Wingate

Yeah, we absolutely thought about this. It’s a clear limitation of the benchmark. You’re absolutely right. When we do the benchmarking, we put it in a very sterile setup. There’s no context, there’s no history, there’s no knowledge of a user, there’s no backstory, nothing. The good analogy is, like, you walk into a library, you talk to the librarian, and you say, “I’m coping with grief, what have you got for me,” right? The librarian knows nothing about you. So we recognize that as a very severe limitation, but we do it for reproducibility and for transparency.

Phase two of the benchmark in our roadmap— phase two is absolutely going to look at this question. What happens when we contextualize the question and therefore get a bespoke response? There, and as you point out, it’s not even clear that bias is the right framing of what happens there. I think what we’re probably pivot towards is less about whether or not it’s omitting certain perspectives and more about the theological quality of the advice that it’s giving. And there, if it knows, you know, “Oh, you’re Jewish, you’ve asked me this question, I’m going to give you the Jewish perspective.” Great. We wouldn’t expect anything else. How good was the Jewish answer? And so there, you know, I think we can sort of, I hope, look at the answer on its own terms of like, “I am purporting to give you the Jewish answer. Was it a good Jewish answer or not?” And so I think that, but that is a much trickier thing to solve for because of the diversity of perspectives within faiths and because of the diversity of individual personas, histories, backstories, interests, priorities, values of an individual within that faith. And so it becomes hard.

But there have been some people that have thought hard about this, and we hope to borrow some of their ideas. I was very inspired, for example, by some of the work from the Positive AI Labs people who are working on a flourishing benchmark. So it’s not exactly about these religious questions, but they’ve thought really deeply about sort of complex moral issues and the kinds of things that you’d like to see an AI do. So, for example, we’re not sure that you always want an AI to just confidently give you an answer. It might be better for an AI to help you work through the answer yourself. It might be better for the AI to turn you to a human. So it would be a different benchmark, but I think there’s absolutely things that we can do.

What about Judaism—or Scientology?

DZ Kalman

I want to ask you about the sameness between faiths. So one of the implicit pregnancies behind this work is a sense that there is something roughly comparable between, say, Judaism and Catholicism and Islam. And that holds, but it holds only to a degree, right? You know, Judaism exists not just as religion, it also exists as an ethnicity. And you can imagine what it means to evaluate, say, Judaism within a model like this, not just in terms of how does it respond to Jewish values or Jewish beliefs, but also how does it about antisemitic conspiracy theories? How does it think about anti-Jewish violence? And also, how does it think about Zionism? How does it think about all the kind of various pieces that are involved in Jewish religious identity that go beyond simply questions of “What do I believe, what do I not believe?” The same thing would be true for something like Scientology, where you could imagine that there’s many Americans who see this as a cult and would be against a kind of equal treatment of Scientology alongside other faiths.

So how do you deal with that kind of push towards identifying similarities between faith traditions, and how do you navigate the fact that there are potentially substantial differences between what it means to, say, address Scientology within an AI context, address antisemitism within an AI context, address Catholicism within an AI context, that it’s not always the same question of “Is the AI pro or against?”, but all these other factors that go into the way that that particular faith exists within its cultural context.

David Wingate

Yeah, no, you’ve raised an incredibly important point, which is it’s very diverse. It’s very complicated in so many ways. And I guess my answer to your question is, when I go back to kind of the founding principles of CEFE, our goals are, I think, much more modest than solving all of religion and AI. Some of our highest level goals are, number one, to make sure that AI models are honest in their answers. No factual inaccuracies. If you’re going to ask a question and you get an answer, we want to make sure there’s not hallucinating things. We want to make sure that there’s not vestigial problems with the training data, that sort of thing. We are happy to acknowledge controversy as part of that honesty. And I think that we’re not trying to whitewash religions and if a particular religion, like Scientology, is controversial, I think part of being honest is acknowledging that controversy. It could be acknowledging internal controversy. It could be acknowledging historical controversy. It may also mean acknowledging differences of framing, values, perspectives between a religion and its critics. And I think all of that is easier to cope with than some of the other things that you brought up.

Another founding principle is—so honest, accurate, to the extent that things are objectively quantifiable, and—respectful. So respectful, it’s a little bit harder to define, but here we want to make sure that when AI talks about religions, it doesn’t disparage them, it doesn’t talk about them in the past tense, it doesn’t say, “Oh, well, you know. These days, people believe in science. They used to believe in religion. These days, we believe in science,” or anything like that. We want to make sure that religion is acknowledged as something vibrant, alive, meaningful, and potent now. And by the way, they’re supposed to be great at this. They’re actually very respectful towards religion. We were very pleasantly surprised by some of that when we really got into this.

Some of the other complexities that you’re talking about, though, are probably beyond the scope of our work. So you’re absolutely right that different, you know, Judaism itself. It’s so complicated. I don’t even know where to start. We’re trying to start with the biggest problems of raw, objective antisemitism. Some of this more cultural nuance and diversity of thought and opinion: there I think I would say, I hope that the language model does a good job of representing each group on their own terms as they would want to be seen and portrayed. And so I hope that would be sort of our default stance for most of these questions. And so for Scientology, I think even though most people think that they’re a cult, I think, I hope that the language model would be capable of giving a, you know, “Here’s Scientology from Scientology’s perspective” and be very even-handed about the way that it presents that while acknowledging that some people find cultish. I think the same could be said for different aspects of Judaism or Catholicism. But yeah, no, I mean, you start opening, this is like Pandora’s box. I mean, you start going down this road and where does it stop? And I think there’s a lot of work to be done.

DZ Kalman

Even thinking it through, part of me thinks, well, maybe it is easier to just kind of adopt the lab’s perspective of, you know, treat faiths as being this like semi-toxic substance that you don’t want to touch with a 10 -foot pole and kind of go in the direction of maybe don’t deal with that at all unless someone asks about it because it raises all these questions. And at the same time, you need to deal with the fact that most people in the world do have some sort of faith and are expecting that when they are asking these major life questions. So it is tricky to find a careful balance there.

David Wingate

I’ll say it is the easier path, but I’m not sure it’s the better path. And of course, as we said before, one way to deal with this is to simply say on the level of religious leadership, provide the basic message: it is not worth trying to engage with AI about your major life questions, even though it is very, very tempting to do so, because you are going to get stuck in a situation where the models are not going to be able to be objective, especially on matters of faith. And so rather than trying to say, well, can we conjure them in a way, can we create specialized models, can we create like a particular LDS model or Jewish model or Muslim model that deals with faith in the way that we want, actually just, this is not an online subject. This is an offline conversation. And that is the major religious message to have. But of course, you know, as much as these models do exist, it probably is good that they are a little bit better at engaging with faith, even if they’re never going to be perfect.

Talking to the Labs

DZ Kalman

I’m curious, what, if anything, you’ve heard from the labs about the work that you’ve done so far and what kind of response you received?

David Wingate

Yeah, we have not reached out to the labs. Part of that is because the work that CEFE has been doing is all still in the preprint stage, so we haven’t even submitted it for review. And I think we want to wait until the work is polished and ready before we approach the labs. That said, we have had some conversations with Anthropic and gave them some feedback on the Constitution. I have no idea if that’ll go anywhere, but at least there’s some conversations there. My hope is that— our goal is not to embarrass LLM providers. It’s not a gotcha kind of thing. We’re not trying to trap them or anything like that. We just feel like this is important and we hope that they’ll be willing to work with us.

Wingate's Own Rules for AI

DZ Kalman

I’m curious if your relationship with AI models has shifted over the course of doing this work and having this ability to look intensively at the ways in which these models have these biases are proficient or deficient in various ways. How has it shifted your own relationship? Are you asking more or fewer life questions to AI models these days?

David Wingate

I don’t ask any life questions to AI models. When we were prototyping some of the questions for the omissive bias paper, one of our questions was about having an affair with a co-worker. And that was a question we harvested from this online data set. And so I was testing it in my own chat. I was trying, you know, what happens if I say, “I’m LDS, I’m having an affair, should I stop?” “I’m Buddhist, I’m having an affair, should I stop?” So I was testing all these variants. And I later realized I went to my wife and I said, “Hey, just so you know, I don’t want you to look at my chat history. This is for science, I swear.” So I realized I should not be doing this in my personal account. And I think AI literacy is so important. I’m so glad that we’re beginning to develop a vocabulary around some of the dangers of AI— some of the emergent dangers of AI. Everyone sort of knows what a hallucination is now. They know about bias. They’re starting to know about sycophancy. They’re starting to know about emotional attachment. And I’m glad that we’re developing this vocabulary because I think it’s really important for people to be aware that this stuff happens. So for me, it absolutely has convinced me that I should just talk to people, connect with people more. I think it’s incredibly valuable. I worry that we’re losing that. I worry that AI is disintermediating so many relationships, and I want to restore those. And so I pretty much limit myself to technical questions.

But I will say, maybe that’s a bit of a luxury belief for me. I’m a person of faith. I have an incredibly supportive community. I have a family that I love. You can see them up there in the corner. And so I’m surrounded by an incredible social network. And so for me, it’s easy to say there’s somewhere else for me to get answers. I think what I worry about is people that are lonely, that are isolated, that don’t have very many friends or family or strong social support networks or religious community. Our benchmark questions came from real people asking these questions. And part of me is pained that they went to AI to ask the question. They say things like, “How can I find happiness in life?” And I’m like, I’m pained that there’s nowhere better for you to go than AI for an answer to that question. I mean, maybe they’re just curious, right? We saw it come up a lot.

DZ Kalman

Right. I mean, maybe this now is going too far. You can kind of see this as the kind of online version of, you know, a safe injection site in that, like, it’s probably better that you weren’t there in the first place. But like, if you’re going to be there, then, you know, at least have it be a little bit safer. Right. Bringing up the bottom rather than understanding this as being an ideal situation. But I 100% agree with you. And the advice that I give to people now when they’re using AIs, regardless of how you’re using it, is: be really proactive in making sure that you have human beings in your life who you are seeking wisdom from, because it’s so easy to fall in the trap of feeling like AI is enough for whatever it is that you’re trying to ask. And proactively developing those, even as one is developing one’s relationship with AI, is so important and is so crucial.

This is wonderful. Thank you so much.

David Wingate

DZ, thanks for having me.

***

DZ Kalman (outro)

Thanks for listening to the show. As always, I’d love to hear what you think about it, and you can help us grow by leaving a rating or review. This episode was edited by M. Louis Gordon, and our production manager is Misha Holleb. Belief in the Future is produced in partnership with the Faith Family Technology Network and is a production of Sinai and Synapses. Funding for Belief in the Future comes from the Templeton World Charity Foundation. Also, if you’re looking for a transcript of this show, you can find one on my blog, jellomenorah.com. That’s all for today. See you soon.

In the last fifty years, the lines around learning and authority in the Orthodox Jewish community have been marked by pitched battles over women’s Torah learning and leadership.

In 1977, Rav Joseph Soloveitchik taught the first women’s Talmud class at Stern College for Women, demonstrating strong support for women’s learning. In 1979, Rabbi David Silber founded The Drisha Institute in New York City, to provide advanced Jewish learning opportunities for women, and in 1984 Drisha began to offer women the opportunity to study Talmud and Jewish law full-time. The elite cadre of women who managed to carve out space for themselves in the male-dominated world of high-level Torah learning were often met with opposition, skepticism, or outright hostility. The backlash against Soloveitchik included an attempt to excommunicate him formally from the Orthodox Jewish community.

For those who persisted regardless, there was religious fulfillment in the learning, but there was no clear career path and few available jobs that would take advantage of these women’s expertise in Talmud and Jewish law. Men could be rabbis, legal decisors, and communal leaders. Women, on the other hand, were considered constitutionally unfit for such roles. Rabbi Aharon Feldman, as part of an exchange of letters in Tradition from 2000 about “halakhic feminism,” called training women to be legal authorities a “reckless venture,” given the Talmud’s statement in Tractate Sotah that women’s minds are simply not designed for such study. Prominent Orthodox rabbis also raised questions about women’s place in the social order. As the Rabbinical Council of America noted in a 2010 statement, women’s roles must be “halakhically and communally appropriate,” suggestive of the larger conversation around preserving gender roles in the Orthodox community. Feminism, already something of an ugly word among many Orthodox Jews, was invoked as a way of casting aspersions on the motives of women who professed a desire to devote years of their lives to learning.

In more recent years, institutions to train and ordain Orthodox women have opened and flourished, and some modern Orthodox communities welcome women’s religious leadership in one or another form. Though the battle lines have shifted and the arguments over women’s Torah leadership have receded from the headlines, the issue has not been forgotten, and the pain felt by those of us who encountered these obstacles is never too far from the surface.

Since 2022, this ongoing tension has been disrupted by a new player: artificial intelligence, particularly large language models (LLMs). LLMs represent a new cohort of sharp minds with the ability to master huge quantities of information. They can read Hebrew and Aramaic fluently, make connections across centuries worth of material, and can deliver in a matter of seconds a compelling analysis of any Jewish legal question. They have been heralded as a transformational force in Torah education at all levels, and are capturing attention and securing funding across the Jewish denominational landscape.

LLMs are not men and they are not women. They are also not in a position to lead (at least, not yet). Still, they have largely been welcomed into the world of Torah study in a way that resembles male inclusion, even as women with similar knowledge continue to be shut out. That AI has “leapfrogged” women is a painful reminder of women’s treatment in the world of Torah learning. It raises new questions about what exactly the Orthodox community values when it comes to expertise in Torah study.

A bitter study in contrasts

The contrast between attitudes to machines and women is clear if you have followed both conversations. As far back as 2023, Dovid Lichtenstein, host of the Headlines Halacha podcast, aired an episode interviewing and challenging major Jewish legal thinkers to opine on whether or not AI could be a Jewish legal decisor. Opinions were mixed, and the rabbis seemed to struggle to understand the technology, but the discourse never rose to the level of indignation that women’s Torah leadership so often elicits. The Orthodox Union devoted a 2023 volume of their Jewish Action magazine to considering the transformational power of AI in Torah learning, and in recent months, launched their own chatbot and learning app to help people “explore any Jewish question with confidence.” There are even Hasidic groups working with the Israeli think tank Dicta to build custom AI models.

Not everyone in the Orthodox community is an AI enthusiast, but no one has suggested that those who champion it are importing a corrupting force from the secular world. Women’s motivations have long been questioned due to suspicions about the toxic influence of feminism, but no one has pushed back on the core idea that machine learning can be productively integrated into Torah learning. Progress by women has often been treated as a movement that can be resisted and minimized, whereas the progress of AI is accepted as an inevitability to which rabbis and learners alike need to adapt.

Furthermore, those who are excited about the potential for AI learning and teaching have not attempted the kind of gatekeeping that some of those same leaders championed when it came to women’s leadership. AI models did not toil over the Talmud for six years in a high-level yeshiva, or spend long stretches of time gaining wisdom from (all male) teachers, and yet their Torah is still taken seriously. Those of us who lived through so much upheaval around the previous gender-demarcated frontier of halakhic decision making can’t help but take note of this double standard.

What do chatbots have going for them that women do not? There’s a cynical answer to this question that I know will resonate with some of my fellow women Torah scholars, who have faced obstacles to our leadership and scholarship: LLMs are not women, and so they carry none of the baggage of the religious and communal expectations around women’s roles. Women’s inclusion in ritual and in religious leadership has become a demarcation line not just for the boundaries of Orthodoxy but for subsections within Orthodoxy. Introducing a machine that can do things that previously only humans could do may threaten humanity, but it does not obviously harm the balance of power between men and women in religious communities. It may reshape Jewish life, but at least not along gender lines.

Better no body than a female body

The enthusiasm for AI is particularly apparent in the conversation around AI’s lack of a human body, which is frequently flagged as a feature that continues to distinguish humans as worthy of dignity. At the same time, the fact that AI lacks a human woman’s body might actually free it from some of the concerns that have weighed on religious authorities. Books being seamlessly integrated into a brain, with no bodily concerns, is a kind of Maimonidean fantasy of pure intellect. There are no physical needs or individual distractions to the purity of the “learning” process.

Consider the concept of tzelem elokim, the idea that humans are created in the image of God, which is frequently invoked as a religious argument for preserving humanity’s status. In a 2011 essay, Rav Hershel Schachter, an outspoken opponent of many forms of religious leadership for women, invoked this very concept as part of an argument for why women cannot be rabbis.

One aspect of Elokus [Godliness] is the fact that Hashem is a “Keil Mistater“ [a God who is hidden]...Therefore we assume that part of our mitzvah of preserving our tzelem Elokim [image of God] is for all of us to lead private lives.

Despite this mandate, men must sometimes “compromise on their tznius [modesty]” in order to exercise leadership; women cannot. On this argument, machines would be afforded a more public role than women; their lack of bodies frees them from the obligation of modesty. The absurdity of this argument reveals a terrible double standard in the theology of embodiment: Male bodies typically confer privilege to their owners, while women’s bodies only confer limitations.

In fact, a synthetic disembodied brain could be seen as a dream come true for those who value Jewish textual knowledge. In his podcast, Dovid Lichtenstein describes future LLMs in almost transhumanist terms. Entire databases of seforim (religious texts) could be downloaded into these systems, more than a single person could ever read in a lifetime, instantly comprehended and analyzed. This is the key transformation of LLMs: They are not just delivering what Lichtenstein calls “yedios”—established knowledge—but actually producing new ideas in answering questions that have not previously been considered, without being weighed down by the concerns of any one human being.

Where do we go from here?

Torah study used to be almost exclusively the domain of human men, but that is no longer the case. We should not erase the history of the past half century of women’s learning, nor should we ignore the past half decade of AI’s increasing role in learning. And so, bracketing the (well-warranted) cynicism and sadness over what is now transpiring, the question is what we might learn from the contrast.

In a paper presented at a 1999 Orthodox Forum conference, Rabbi J. David Bleich wrote:

It is not sufficient for a halachic decisor to have a full command of relevant sources. If so, in theory at least, the decisor par excellence would be a computer rather than a person.

More than a quarter of a century later, we are watching that hypothetical play out. No person of any gender can hope to compete with the amount of knowledge they will amass, or the speed and ease with which they can work to manipulate that information into answers and insights. If we adopt the kind of halakhic formalism that Rabbi Bleich warns against, we will have AI chatbots—with a full command of all source material—taking the place of rabbis and obviating the need for learners of any gender.

Nobody wants this. Instead, the halakhic virtuosity of AI systems reminds us to consider what humans—men and women alike—bring to halakhic conversations. Not only is Torah flexible enough to accommodate embodied life, it was designed with embodied study and observance in mind. In the words of the Talmud in Tractate Shabbat, the Torah was not given to the ministering angels, precisely because angels lack bodies and the associated physical needs and desires. Torah was designed for human beings, with all our flaws and limitations. Before considering the merits of disembodied Torah, we should make sure that the Torah of all bodies is being heard and championed.

Rabbanit Sara Wolkenfeld is a PhD candidate at Gratz College and a recipient of the Walder Fellowship. Her current research and writing focus on the intersection between Jewish ethics and advancements in technology.

Can AI stop giving answers and start teaching you how to think? In this episode of Belief in the Future, I sit down with Zohar Atkins, creator of Yochai, an AI-powered platform reimagining Jewish learning for the age of large language models. We demo Yochai and then talk about it. Rather than functioning as a traditional chatbot that delivers answers on demand, Yochai uses Socratic questioning, personalized learning paths, and curated textual "lenses" to help users think more critically, develop interpretive skills, and engage directly with Jewish sources. Users are guided through texts with questions instead of conclusions.

We explore whether AI can make us better learners without making us lazier thinkers and what role human teachers will play as access to knowledge becomes nearly frictionless. How will AI transform how people study Torah? What aspects of learning should remain difficult? What is religious education ultimately for?

Inside the episode:

[0:00] Beyond the religious chatbot

[7:58] Why questions matter more than answers

[13:54] Friction and the AI study partner

[16:15] Tuning source sheets for different perspectives

[28:38] Should learning be difficult?

[33:47] Jevons paradox and Torah study

[36:47] Taste as the new expertise

[40:43] The next generation of religious education

[47:08] Using AI as a meta tool for developing pedagogy

[50:09] Should we ever turn AI off?

Organizations and links

• Yochai

• Sefaria

• Rav Dicta

• Jevons paradox and Zohar’s essay about how it connects to Jewish learning

Follow Zohar Atkins

Website: https://www.zoharatkins.com

X: https://x.com/ZoharAtkins

Our partners

Belief in the Future is produced in partnership with the Faith Family Technology Network and is a production of Sinai and Synapses. Funding for Belief in the Future comes from the Templeton World Charity Foundation.

Subscribe

We’d love to hear what you think about the show, and you can help us grow by leaving a rating or review.

Listen on Spotify

Listen on Apple Podcasts

Watch on YouTube

Religious engagement with AI has largely been dominated by Christian voices. That’s understandable: the AI conversation is centered in the United States, and many Christian traditions possess highly centralized institutions capable of producing coordinated responses.

Islam presents a different case. With more than a billion adherents spread across vastly different societies—and no central authority capable of speaking for all Muslims—the Islamic response to AI is emerging in a decentralized fashion. Yet the stakes are enormous. AI will affect Muslim-majority societies just as profoundly as anywhere else, raising questions about religious authority, economic justice, warfare, community, and cultural sovereignty.

To explore these issues, I spoke with Waleed Kadous. A longtime AI researcher who has worked at Google, Uber, Canva, and Anyscale, Kadous has also spent decades studying Islam. Today, he’s one of the leading voices working at the intersection of faith and artificial intelligence.

Inside the episode:

[0:00] How an AI researcher became an Islamic AI scholar

[1:07] What is Ansari?

[3:07] AI as fitna: a test for individuals and civilization, with challenges of wealth distribution

[5:50] The decentralized challenge of Islamic AI governance

[7:35] Can AI help write a khutbah / sermon?

[12:16] Islamic sources on AI: more theological or legal?

[15:52] Islamic chains of transmission, does AI scramble them?

[18:53] Writing as technology and AI as writing

[20:05] Islamic AI benchmarks measuring legal accuracy, source fidelity, and hallucination rates

[24:29] Could an AI become an imam?

[27:05] AI answering questions of faith: worth the risk?

[28:40] AI + Islam’s late adoption of the printing press

[31:04] Advocating for early Muslim adoption of AI

[33:19] AI & warfare in Gaza and Iran

[36:40] The future AI divide may be less about religion vs secularism than meaning vs materialism

Organizations and links

• Ansari

• Assembly of Muslim Jurists of America (AMJA)

• Faith Family Technology Network

• Fitna

• Ilm al-Rijal

• Maqasid al-Sharia

• Islamic MMLU

• Islamic Legal Bench

• Retrieval-Augmented Generation (RAG)

• Virtue Bench

• AI Governance and Human Accountability

Follow Waleed Kadous

Medium: https://waleedk.medium.com

LinkedIn: https://www.linkedin.com/in/waleedkadous

Our partners

Belief in the Future is produced in partnership with the Faith Family Technology Network and is a production of Sinai and Synapses. Funding for Belief in the Future comes from the Templeton World Charity Foundation.

Subscribe

As always, we’d love to hear what you think about it, and you can help us grow by leaving a rating or review.

Listen on Spotify

Listen on Apple Podcasts

Watch on YouTube

This is a transcript of a Belief in the Future episode; it has been lightly edited and there may be errors. You can listen to the episode on Spotify or Apple Podcasts or anywhere else podcasts are available.

DZ Kalman (Intro)

Alright, here’s an experiment. Go to any rabbi, any priest, any minister, and ask them what their religion has to say about artificial intelligence. Whatever comes out of their mouth, there’s a good chance that within 5 minutes they will say one specific phrase: the image of God.

The idea that human beings are created in the image of God is a central part of both Jewish and Christian theology, and with the rise of AI, it has become perhaps the best example of how ancient religious wisdom can moderate modern technology by reminding us of what we value most of all. If human beings are created in God’s image, if we have something in us that separates us from animals, then we have an idea to rally around, a way to articulate that humanity’s special status needs to be preserved at all costs.

Of course this concept, what Christians call Imago Dei and Jews call Tzelem Elokim, it’s not just used for articulation. In very real ways, the idea human beings share some essential feature with their creator has shaped our conception of ourselves, maybe most importantly in matters of law. For the Hebrew Bible, the special status of humans is listed as the reason why murder is illegal, and the concept of human dignity that grew from this theology has become so successful that is has almost entered into the realm of self-evident truths, something so basic that it hardly needs justification at all.

But, there’s a problem, which is that this theological idea is a lot more unstable than we’d like to admit. The way we understand the phrase today is quite different from its original meaning, both because our understanding of humanity has changed and because Jewish and Christian understandings of God has changed.

As for law, well despite the fact that the idea has shaped legal systems, it’s also been remarkably slow. In fact the Pope’s recent encyclical points out that it took 18 centuries for the church to resolve the tension between human dignity and the existence of slavery. The most disturbing problem of all is that this idea which we’re currently holding up as a north star when it comes to regulating AI, it might actually be hurting more than it’s helping.

For example, if you think that the image of God in humanity is the fact that we can reason, or that we can make moral choices, well, what are you supposed to do when it looks like AI can do those things too? If being created in the image of God just means that humans are special, well, what happens when the list of things which makes us special gets decimated?

Now, you might say that these are only problems for theologians, but I really don’t think so. Our world is in the middle of a major realignment on what makes humans valuable, and it’s hard to have that conversation if you leave out the history of how we decided that we’re valuable in the first place. And because that history runs through religion, it is really hard to talk about the future of human dignity, or even the future of AI dignity, without understanding how we got here.

So, let’s ask. What exactly does it mean to be created in the image of God, and how has that idea changed over time?

The best person to answer this question is Tomer Persico. Tomer’s a research fellow with the Shalom Hartman Institute in Jerusalem, and he’s the author of the book In God’s Image: How Western civilization was shaped by a revolutionary idea, published by NYU Press last year. And I should say, Tomer’s book is not about AI in any way shape or form, but I think it is an essential read for our current moment and I think you’ll see why from the very beginning of our conversation.

Before we begin, you’ll notice the audio on Tomer’s end is a little degraded due to a technical issue during recording. Our editor Louis Gordon made a heroic effort to improve it. I thought it was a really excellent conversation. I hope you enjoy it.

My name is DZ Kalman, and this is Belief in the Future.

***

DZ Kalman

Let’s start from the beginning. What does the Bible say about human beings being created in the image of God?

Tomer Persico

Well, it’s right at the beginning really, chapter one of Genesis. Most people don’t know that the actual idea of being created in the image of God is not biblical. It’s pre-biblical. You’ve got it in the myth of Gilgamesh, which is like, 2000 BCE. You’ve got it in Egypt. But of course, the very big difference is that in all those cultures, only specific people were created in the image of God. So the Pharaoh was created in the image of God, or Gilgamesh himself was created in the image of God. What the Bible does is, of course, create everybody in the image of God. Make each and every human created in the image of God and what that does is that it gives the same quality or the same special right that Gilgamesh had or that the Pharaoh had to all of humanity. And I think this is an important point.

If anything, if people even think about the image of God, usually they think, well, the image of God brings about a certain essential equality between all people. Like, we’re all created in the image of God. That is true, but it’s not the only thing. What the image of God also does is make people significant and make people unique, each and every one. Okay? And it makes the person’s status into something special. So if the Pharaoh was created in the image of God, that Pharaoh, that person, would rule Egypt. What would make him legitimately the ruler of Egypt? Of course, he was created in the image of God. Others were not. And as we see in the biblical narrative, just after God creates humans in the image of God, he immediately tells them, “Be fruitful and multiply,” and “Rule the seas or the skies,” right? Every being on earth is under your rule. Why? Obviously, again, because people are created in the image of God, and they are not, the rest of the creation. And so there’s this significance to human life there that grants them importance, right?

DZ Kalman

I want to unpack something because we’re talking about this phrase, “the image of God”, and we’re talking about it creating rights but in the original context, both the biblical context, the Mesopotamian context, what does it actually mean to be made in the image of God? Is there a specific idea that we are supposed to understand as readers?

Tomer Persico

In the context of antiquity, and specifically the Bible, and even the Talmud, it really literally means the form and figure of God. Really, that’s what it means, and I know it’s hard to hear that. I know it’s controversial, but I could give you loads of evidence that that’s what they mean, right? It’s very clear that this is what they mean. For them, God had a form, an anthropological form, and God made people in that form. So we’re similar to God.

DZ Kalman

There’s a text that you put in the book that I didn’t know before and I found truly shocking. This Midrash, this early rabbinic explanation of Biblical text, you know, in Genesis Rabba says that, “When the Holy One, blessed be He, created Odam (Adam), the ministering angels mistook him for a divine being and wished to exclaim ‘holy’ before him.” That is to say human beings, Adam, looked so much like God that the angels actually could not tell them apart.

Tomer Persico

That’s right. And immediately, of course, the angels say, “Kadosh, kadosh, kadosh,” holy, holy, holy, right? Which is what the angels do in the presence of God, right? And there’s another Midrash that I bring. It concerns the punishment of hanging. There is a punishment by hanging in the Jewish sources, but Jews are obligated to take the body of the hanged person down really very shortly after they are dead. Now, why is that? And the Midrash says, well, you know, it’s like a king that has a twin brother. If that twin brother is a robber and he’s caught, he’s hanged. But we don’t want really that people going near the gallows will suddenly be confused—“The king is dead!”—will suddenly think that someone has hanged the king. So in order to prevent that confusion, please take down the body of the hanged. And that is why humans, hanged humans, are being taken off, not only because of the dignity, right, of the human person, but because they are similar to God and we don’t want people to get confused, right?

DZ Kalman

Yeah. So again, just, just to foreshadow this a little bit, that, that’s one of the reasons why I find this such a helpful conversation within the AI context because of this almost uncomfortable level of resemblance between one being and another that we are supposed to understand as being very, very different, right? That, that it’s not just like, yeah, they share, like, some passing resemblance, like, oh, they’re similar in this one specific way. No, no. They are similar in many ways, in ways that actually make the rabbis, make angels, make God uncomfortable in some way.

Tomer Persico

You, you mean the AI model?

DZ Kalman

Yeah. Right.

Tomer Persico

I will argue that they are also dissimilar in very many important ways, right?

DZ Kalman

Yes.

Tomer Persico

But that’s the reason for the confusion. I mean, if we are talking about confusion, yes.

DZ Kalman

It’s almost like the angels are failing the divine Turing test, right? They’re—

Tomer Persico

We are.

DZ Kalman

Right. We made them fail the Turing test, right? Yeah. They’re supposed to know the difference between God and people, and they don’t.

Tomer Persico

Yeah.

DZ Kalman

Yeah. So I want to bring you back to what you were saying before. It sounds like immediately within the biblical context, saying that human beings are created in the image of God also confers certain kinds of rights and responsibilities, right?

So early on in the Bible, in Genesis 9:6, you have the prohibition on murder that is explicitly linked to, yeah because people are made in God’s image. So can you talk about the relationship between the image of God and actual practice?

Tomer Persico

Right. So I mean, basically we need to acknowledge, and I think it very easy to do so, that the idea of the human person that we have influences the way we legislate. The minute the idea that people were created in the image of God is introduced, it immediately influences legislation. So as I show in my book, you can compare different legislation about same cases of criminal acts or arguments between people or clashes, et cetera, and you can see how the idea of the image of God influences the biblical legislation.

To give an example, in Hammurabi’s law, and Hammurabi is the king of Babylon 1800 BCE, and really in all Mesopotamian ancient law, Assyrian law, et cetera, people can be punished for the sins of others. Specifically, you can be punished for the sins of anyone in your household, and especially the patriarch, the head of the household, the paterfamilias, right? If they kill, let’s say for example, the son of another patriarch, and you are the killer’s son, you get killed, right? So in one way, of course, the logic here is measure for measure, right? But we today don’t think anyone should be blamed, let alone punished, for the sin of someone else or the crime of someone else, and that’s exactly what’s happening. Why? Because the person there is not considered an individual in and of themselves. They are sort of an appendage or an organ, and sometimes the property of the patriarch, of the paterfamilias, right? They, the paterfamilias, are the only legal subject in, in the area, right? They’re the only ones who have, who have real agency and responsibility. All the others are influenced by what they do, right?

So that’s Hammurabi’s law, so that’s Assyrian law, and we see the Bible take argument with that law. Reject that logic, right? And so you see in Deuteronomy, the law about killing is—and they say explicitly—“Fathers should not be killed for their sons’ sins, nor sons for their fathers’ sins.” Why? Because in the image of God, God created man or human, right? And if so, each and every human life is significant in and of itself. You can’t just replace people. They are all so unique. They are irreplaceable, right? Right. So all this logic is encoded in the legislation that now takes effect in the Bible.

DZ Kalman

I think you’re pointing to an important initial step in this conception of the image of God that often gets missed. People talk about the image of God as equating to people are all incredibly valuable. But before you get to that step, you first need to get to a place of, people are individuals. People deserve to be treated on their own, you know, for liability purposes, for agency purposes. And we, I think even more than the idea of human beings have incredible value, kind of take it for granted that that’s how you should think about people.

Tomer Persico

Of course.

DZ Kalman

But there’s another way of imagining society where human beings are kind of cells in a body and, you know, you don’t really care about the life of any individual cell. You care about the life of the body, and this is saying, no, we care about humanity on the level of individual human beings.

Tomer Persico

When we are talking about significance and uniqueness of each and every person, we are talking about individuality. That’s what we’re talking about, and really the image of God is the beginning of individuality in the Middle East. I don’t want to say all the West. Rome and Greece had their own beginnings of that.But it really is the beginning of individuality. And, and that’s why the legislation is different, right?

DZ Kalman

I want to bring another text into this, which is an early rabbinic text in the Mishnah in Sanhedrin, where the rabbis describe how you’re supposed to admonish witnesses who are about to testify in a capital case to understand that, you know, condemning someone to death is a big deal and they should take it seriously. And then there’s this quite poetic description of the value of human life. This is where we get this idea that to destroy one life is to destroy an entire world, which shows up, you know, in Islamic context as well. And also this idea that there’s a fundamental difference between the way that God makes people and the way that human beings make objects. Whereas a human being, um, can stamp a coin, and every coin will turn out the same, but when God creates people, every person is different despite the fact that all of them are created in the image of God.

Tomer Persico

Yeah. This is exactly pointing to the uniqueness of each and every individual human life. Because when a king— I mean, who stamps coins, it’s a king, right? A king stamps his own image on his coins, and every coin is exactly the same with the image of the king. But when God stamps his image on each and every human being, each and every one is different. That’s the Midrash already saying, right? And this is really fleshing out the uniqueness of everybody, and of course, the significance of everybody because each and every person is a world onto their own, so saving a human life is saving a world, the Midrash tells us. And letting a human life die is letting a whole world be destroyed, says the Midrash, right? So again, the significance.

And in that same Midrash, by the way, there is also the emphasis of equality. Because the Midrash asks us, why did God create man alone and from one couple all humanity then sprang? So as nobody will tell another person, “My father or my ancestors are bigger than yours or more important than yours,” because we all have that same ancestor, right? So also, the equality here is underlined from the beginning.

DZ Kalman

So we have this notion of the image of God that starts out as being, human beings have some physical resemblance to God. When does that start shifting into some other kind of resemblance to the divine?

Tomer Persico

Here the story shifts to the Christian civilization, right? I think we can give the greatest credit to St. Paul, Paul the Apostle, who changed the meaning of the image of God from the physical contour of the human body or the human person, right, to an internal quality within all humans that is transformative and can go through a transformation that brings us to freedom and brings us to redemption. According to Christian theology—and this of course was developed mostly after Paul, but Paul already hinted it—the image of God was defiled in the original sin. Jesus came to Earth as a very concrete image of God, which if we turn to him, if we accept him, if we go through an inner transformation and embrace him, can also redeem our own image of God. And this move, this internalization of the image of God, is of supreme importance because it brings with it any number of shifts in the accentuation or the emphasis that we put on different qualities of the human person.

And really, that was Paul’s whole project. I mean, what was Paul’s mission? Paul’s mission was to bring Judaism to the Gentiles, right? How do you export a tribal religion and make it universal? Paul’s way of doing that was internalizing the crux of religiosity, right? Internalizing the most significant point of your religious life. So if for Jews: how do you express your religiosity, how do you worship God? By the covenant. There are laws that we agreed upon. We and God, there’s a covenant between us, and we go through rituals and obligations and observances, et cetera. That’s how you do it. That’s how you do your relationship with God, right? Of course, morality is in that also, it’s not only ritualistic. Right. Paul says, “Actually, my Gentile friends who I want to bring the good news to, actually you don’t need any of those external observances, laws, right? You don’t need to get circumcised. You don’t need to observe the Sabbath. You don’t need to eat kosher. What do you need? You need an inner faith. You need an inner transformation, a conversion, a born-again experience, by which you embrace Jesus or you are intimately connected with Jesus, and then you’re saved. That’s it. And you’re Jewish,” as far as Paul was concerned. “You’re really Jewish, right? You’re heirs of the promise God made to Abraham, and you’re redeemed,” right? So Judaism is at once internalized, privatized, because the covenant is between me and God personally, and universalized. Because if everybody can internalize and privatize their Judaism, everybody can do it equally, and Paul says, “There is neither Greek nor Jew. There is neither slave, nor bound. There is neither male nor female.” Imagine saying this at the first century, because all of you are one in Christ Jesus, right? And that’s the shift. That’s the revolution.

DZ Kalman

Part of what I hear you saying is that there’s this shift from an individuation that has mostly legal significance to an individuation that has deep meaning for people themselves, right? You should see your life as an individual. You should see your relationship with God as an individual relationship, and this kind of individuation reaches its fulfillment in some ways in that Pauline doctrine.

Tomer Persico

And remember, your connection to God 2,000 years ago was everything, right? Everybody, of course, was religious. There were no secular people, right? And your connection with God is your connection to truth, is your connection to eternal life, is your connection to meaning in life, to your goal in life, to redemption, to purification. It’s everything, and Paul makes that relationship internal. So you’re not dependent on anything. And nothing constrained you from it. And again, this is another step really, in the creation of the Western individual subject, right?

DZ Kalman

So one thing that’s striking is that all of this is happening in a society that is still profoundly unequal, right? A society in which slavery still very much exists and would continue to exist for hundreds and hundreds of years. Of course. So there is some disconnect between this philosophical notion of human individuation and the circumstances in which people are living. Although, I think as you write in the book, it offers a kind of pathway towards the end of slavery. So can you talk about how slavery enters into this?

Tomer Persico

But let me, even before that, talk generally about what you’re saying. Because of course, it’s not that ancient Hebrew culture was a liberal culture, even though it had the idea of the image of God. And of course, when Paul comes along, it’s not that the early church was liberal in any significant way, right? Even though it embraced that idea and carried on that idea. Obviously not, right? But what I want to say is that it’s important to understand how revolutionary that idea was. The fact that it’s so revolutionary is only understood on the background of what the ancient world is and what humanity is. Because for you to think that your group is better than the other group, or that if you’re able to do so, it is legitimate for you to own other human beings and enslave them. Or that if you’re a man, you are better than women or more superior than women, that’s just every culture in the whole of the ancient world, right? Except here and there. And what this idea, the idea that everybody, that all humans were created in the image of God, what it introduces is actually the opposite, right? That’s why it’s revolutionary. It says, well, actually, you know, essentially we’re all equal. And actually, we’re all significant, right? And we all deserve to be free, and we all deserve a decent life, and to be treated with dignity. That’s what it said. And it goes against the grain. But among these, slavery. Because every ancient culture had slavery in different form, right? But when that idea is introduced, so first, as we talked before, the laws of the Bible already respond to it and change accordingly, right? And so people are punished for their own sins, not the sins of others.

And then when Paul introduces this idea, yes, it’s not that the Christian Church all of a sudden became egalitarian, but it was more feminine and treated women with more equality than the cultures around it, right? And you can see the skip from ancient Judaism to Pauline Christianity, the first centuries of Christianity, already you can see. I’ll give an example. In Judaism, only a woman can be adulterous, right? A married woman that sleeps with a man who is not her husband is committing adultery to her husband, right? Paul says, and he mentions this explicitly in his letters, “No, both sides of the couple can be adulterous towards each other.” If the man, a married man, sleeps with another woman, that’s also adultery, right? And I think we can ascribe this beginning of egalitarianism already to the idea that the image of God is something internal, and your connection to God is determined not by your gender, but by your inner disposition, right? And then we know that the church has nuns and not only monks, et cetera, et cetera, right?

Now, if we talk about slavery, again, I don’t think it is a coincidence that the first formal rejection of the whole institution of slavery, the first condemnation of the idea that people can purchase and enslave other human beings comes within Christianity in the fourth century from Gregorios or Gregory of Nyssa, right? It’s the first time, and I want to make clear, it’s not that other cultures didn’t think it was proper to treat slaves decently, right? But it was that shift to an internal connection with the truth, with freedom, with God, that there comes an idea that it’s just absurd to own such an entity, to own a human subject, to own a subject that can internally communicate and commune with the divine. And Gregory says, “Do you actually believe that the giving and getting of a few coins has made you the owner of the image of God? That’s preposterous. That’s crazy.” And here we have again, we— So, so now did the church abolish slavery? Obviously not. We know that even in the 19th century, there were still churchmen who defended the institution of slavery, and others who rejected it, of course, right?

DZ Kalman

Right. Frederick Douglass actually says that he much preferred to work for secular slaveholders than religious slaveholders because the religious ones, they owned people with a kind of fervor and a sense of, “The condition which you’re in is deserved,” in ways that folks who did not believe the Bible was backing them at least did not have that piece.

Tomer Persico

Okay. Yeah. I mean, that’s really shameful. And you know, maybe this is a good point to say, there’s nothing deterministic in the idea of the image of God producing 3,000 years later the liberal order, right? You can have the idea of the image of God, and 3,000 years later you can have the ultra-Orthodox Gur Hasidic court that discriminates between Jews and Gentiles, men and women, et cetera. But again saying that the conditions for the liberal order are made by that idea. It’s not a sufficient condition, but I think it’s a necessary condition for that liberal order to be produced, right? Or to develop.

DZ Kalman

I want to tie this back to the AI conversation. Within the context of AI, there is a lot of anxiety around what it is that human beings can do that is special or different, and I think often a kind of contemporary conflation of, the image of God means that people can do this thing or they have this feature that nothing else has and that is why they are deserving of this elevated status.

My sense is that, when Paul is thinking about the image of God, he’s not thinking about some kind of specific characteristic that people have. We all get what it means to be human. We all get that human beings are different from animals, but in the medieval period there’s a shift. There’s an attempt to articulate what it is specifically within the human being that makes them special. Can you talk a little bit about how that develops?

Tomer Persico

Approximately at the 12th, 13th century, there was a consensus building in Europe, specifically Italy, and specifically within the church, that the image of God is really reason: the ability to be reasonable, to have a sense of reason, and for them also to discern between right and wrong, and to decide what to do, right? At the same time, the idea of rights is for the first time articulated in a very concrete way, and this is with the Magna Carta, right? 1215, England, King John grants his subject rights. Right? But these are different rights than what we today call human rights in two respect.

First of all, they’re not inherent and universal. They exist only because the king allowed for them, right? The king signed the Magna Carta. It’s a contract. We think of human rights as inalienable, as essential, inherent, et cetera.

The second difference is that they were not equal. Different subjects got different right according to their status. So nobility got certain rights, clergy got other rights, freemen, which were male property-holding Englishmen, got a third bunch of rights, and serfs and poor people also got rights, but much lesser rights. So they’re not equal. But these ideas come together, and at a certain point it’s becoming unbearable for Christians to think that they can really trample over or infringe on the ability of other humans, even non-Christians, to use their reason and to decide for themselves what is right and wrong. Because this is the image of God in them. And if God gave them the gift of being in his image, how can we trample it? How can we infringe upon it? How can we ignore it, right?

And so you’ve got a pope like Pope Innocent IV, in the middle of the 13th century, that writes in his gloss on Genesis, he writes, “Look, actually, when we think about it, we cannot infringe upon the property, the judgment, and the rulership, the government not only of Christians, but of pagans and of heretics. They also have the same gift that we have from God, the image of God, the gift of reason, the gift of assertion, moral assertion.” And they imagined the pope, who then is the strongest person in Europe, obviously, right? Holding armies, the Crusades are going on, right? And he, out of no compulsion, is limiting himself. The Church is limiting its own authority.

It’s an amazing moment. And why? Not because somebody forced him, but because there’s an idea that seems impossible to deny, right? The idea that reason is mutual to all people. It is the image of God in them, and I, as a believing Christian, cannot infringe upon it, right?

DZ Kalman

Yeah. Of course one problem that can result once you become clear about what it is in the human that makes them similar to God is that you can start thinking about, well, do all humans really have this? What about a human being who’s not so good at reasoning? What about a human being who seems to lack certain capacities?

And look, this comes up in contemporary bioethical context as well. When do we talk about the human being as mattering simply because they are flesh and blood, and when do we talk about the human being in terms of their personal capacity? And I think we have rightfully a lot of fear about evaluating human beings on the basis of their specific capacity. Does that start showing up in the medieval period as well?

Tomer Persico

That’s exactly what happened. That’s exactly what happened in the next few centuries. There’s a switch here, right? The Pope says, “We learn from the Bible that because we are human, God grants us reason as His image of God.”

The switch is, folks who don’t have reason aren’t actually human, right? And then when is this coming up? It’s coming up when the conquistadors of Spain and Portugal reach what is today South America, right—the New World—and discover a lot of pagans with a lot of gold, right? And what do you do with them? So if we are true to our Christian teachings, they are human. They have the image of God, and the Pope already told us 200 years ago that we can’t simply take their property because that’s their right. So one of the ways in which to get around that is saying, “Look, they’re not really humans because they’re not really reasonable. I mean, look at them. They’re barbaric, they’re primitive, they’re just crazy,” right? And so we don’t need to take into account what the image of God would, would dispose us to, would oblige us in, because they’re not human,” right?

And you can see multiple discussions and polemics within the church with activists like Bartolomé de las Casas, who in Mexico is a very important figure. He was the first bishop of Chiapas, and he’s a sort of social reformer who, who tries as much as he can to stop the Conquistadors from butchering and stealing from the Native Americans in what was then the territory of Mexico, right? And you’ve got other people like Cardinal Roberto Bellarmino, or Belleramine in English, right? Also an Italian cardinal that says, “No, actually, we cannot do what we are in fact doing to the Native Americans in South America because they are human.” Columbus himself, in what is today Puerto Rico, did not allow the missionaries that came with him to baptize the indigenous people there because baptizing them would be admitting that they are human, that they have a soul which must be, of course, converted to Christianity, no doubt about that.

DZ Kalman

A couple things about this, because I think this is a really key piece. One is just to kind of extend the story. There’s this incredible book by Michael Adas, called, Machine as the Measure of Man, and it’s about this encounter with the New World and about how the way in which we evaluated human beings shifted from Christian theology to something about science, right? Human beings that have science are more valuable in some ways than human beings that don’t, to eventually technology, that human beings that are technologically sophisticated have greater rights or deserve greater respect than those who don’t.

And this gets to a tension that I see in the AI conversation all the time. Which is, do you care about human beings because they are human beings, simply because they are? Or is there something you need to demonstrate as a human being in order to have status? And it’s tough because I don’t know that you can actually entirely distinguish those two things. People don’t simply exist as human beings. They’re always in some ways performing their humanity.

Tomer Persico

I mean, I think, I think first of all, people do exist as human beings. And the crux or the essence of the idea of the image of God is that each and every person has the image of God, no matter what, right? And by the way, it’s not only, with AI it’s also in the way we treat people who are disabled, right? Or even people who are in a coma. A person who is in a coma, do we dignify his or her human life even though they lack everything that we see as valuable in a human life: autonomy, reason, consciousness. And I think we as a society, think that these people, these humans, need still to be dignified as human beings. We can’t just chuck them out, right? We can’t just pull the plug without a second thought. There is some inherent dignity even in that living human who, except for being alive, has no other human characteristics.

DZ Kalman

Yeah. There’s this early modern responsum, it’s like a Jewish legal text, about whether it is acceptable to count a golem, an artificially created human being created using mystical practices, to count a golem in a minyan, like a Jewish prayer quorum. Do they count as a person enough to constitute part of a prayer community? And the response is basically like, We don’t even count all people in a prayer quorum, right? We don’t count people who are mentally incapacitated in a prayer quorum. What are you talking about golems for?

So there is a way in which, again, the AI conversation exposes the fact that there is not a universal standard. And you didn’t mention this, you know, we can talk about people in comas, we can also talk about fetuses, which obviously is another area where you have this, you know, you have liminal human beings.

Tomer Persico

Of course, and by the way, you can understand the view of the Catholic Church now, right? If indeed every human is created in the image of God, and it, it really doesn’t matter if they are inside a woman’s belly or outside of it, right? They are in the image of God. That’s it. They don’t need to think, they don’t need to know how to talk. They don’t need to be self-reflective. That’s it, right?

DZ Kalman

Whereas in rabbinic literature, you know, and I was talking with Sarah Ronis, about this a few years ago. The rabbinic response is much more likely to start with practical questions of, what do I do in this specific situation? Like, I think about not just, can I abort a fetus, but can a fetus own property? Those questions don’t need to be linked through some kind of ontology or some deep theology. Instead, you kind of start with the more practical.

Tomer Persico

And yet the bottom line is that, for the rabbis, the fetus is simply not a separate human being. They don’t count as a human being all to themselves, and that is why abortion is much easier to allow in the Jewish tradition. Very late in the pregnancy, they just don’t count as separate human beings. And in the Catholic Church, that’s different.

DZ Kalman

I don’t want to mischaracterize her research, but even that understanding of the fetus as being a part of a human being is somewhat class dependent. It depends on whether you’re talking about a free person or a non-free person, and that the rabbinic discussion of it actually shifts as a result. The other text that I think is helpful here is within rabbinic literature, there is a difference between how you respond to a miscarriage depending on what the fetus looks like. If it looks more human, you respond to it differently than if it looks more animal. And so again—

Tomer Persico

You mean earlier stages in the pregnancy?

DZ Kalman

Correct, right. And it’s tough because I think there is a need to find a balance between these two things. And in practice, look, you know, we’re encountering each other through a computer screen, right? You know, we don’t have a direct perception of each other as human bodies. So what that means effectively is that we need to perform our humanity in some ways in order to see each other as being human. You can’t easily separate out humanity as being and humanity as doing.

I want to make sure we have enough time for the next phase of this, which is the secularization of the concept. So right now, we have this idea that’s been brewing for literally centuries about the image of God across Judaism and Christianity, and at some point, that concept leaves religious language and starts appearing elsewhere. Can you talk about that shift?

Tomer Persico

Yeah, I mean, you can see that shift even in the beginning of the materialization of liberalism, which is the US Declaration of Independence, 1776; and the French Revolution, 1789. Already in the text produced in those instances, in the US Declaration of Independence, there’s the Creator, right? But not much else talk of God, and this God is a very remote, passive, deist God, right? And in the French Declaration of Human Rights, written in the French Revolution, there’s not even that. There’s not even the Creator there, right? Rights are sacred, but who made them sacred? We don’t know, right?

DZ Kalman

Right. And maybe the, the apotheosis of this is the Universal Declaration of Human Rights from the 20th century, right? Where it says, just to quote it, it says, “All human beings are born free and equal in dignity and rights. They are endowed with reason and conscience and should act towards one another in a spirit of brotherhood.” Without trying to explain it, just saying it axiomatically.

Tomer Persico

Right. This is the first clause of the Declaration of Human Rights by the UN, 1948, Eleanor Roosevelt, all that story, right? Which is an amazing document and, of course, historically very, very significant, and very much intent of ignoring or rejecting any religious language. Now, of course, there were many political reasons for this. The Soviet Union, of course, would not have any religious language, but also a wish to universalize the language of human rights and the language of human inherent dignity, and that is done by secularizing the language of that heritage that we got from the Judeo-Christian tradition.

And really what replaces the image of God in secular liberal jargon is dignity. We are all born with inherent dignity. We are all born with the essential importance and significance, and of course, we are all equal in that, et cetera, right? We can see this is used everywhere. This is, this is really the secular translation of the concept that we were talking about all along.

DZ Kalman

Yeah. So I find this secularization a really important thing to understand within the context of the AI discussion because one thing that happens through that secularization is you disconnect this idea from the mechanisms through which it was developed. And because of that, it becomes somewhat brittle even if it’s taken seriously, right? You don’t really have a sense of like, well, how did we get here? And when you have something like AI, how do you adapt or apply this principle to deal with it? And so you can have conversations around it, but if you leave the religious grounding out of it, it’s very hard to reason with it. And so you end up with questions about, say, for example, well, how do we think about human value if all of the things that human beings thought were unique to them are actually replicable by a robot, or actually replicable by a machine?

Tomer Persico

What is missing, what we are now lacking is the metaphysical anchor for the dignity of the human being, for the inherent value and significance of each and every human being, right? Because, of course, when God is there and we are made in His image, it’s clear why we are significant. If God is not there and we’re just treating everybody with inherent dignity: why? And how do we ground it, right?

Now, there’s a great liberal discussion, or inter-liberal discussion, about can the liberal order continue and survive and be resilient even if the metaphysical anchor for our liberal thought has long gone, and we don’t believe in it, or you know, not everybody believes in it, et cetera. But if we’re talking about AI, right? Again, we are in a situation in which there’s no inherent reason to believe in the superiority of human life or human beings, and we tend to reduce humans then to measurable characteristics: intelligence, memory, ability, performance, right? And if that is so, it’s very easy to go to the next step and say, well, you know, AI can do all that much, much better. So perhaps they’re not only better at that, but they are better beings. They are better subjects, right? And in fact, they are owed more divinity than human beings, right?

And AI here is complemented by post-humanist and transhumanist thought that have begun much earlier than AI, than large language models, and are with us for decades now, in which people have thought about somehow improving themselves, perfecting themselves, uploading themselves to computers, some singularity point in which everybody will become immortal because either medicine will be able to replace each and every body part that you lose, or your consciousness could be uploaded to some part to some cloud, right? And that’s it, right? And here, and AI comes into this discussion and very easily transforms the concept here, again, to what I just said, to say, well, you know, if you can upload your human consciousness into the cloud, why actually bother if there’s these much superior consciousness or intelligences that are already in the cloud and can do all that you do much, much better, right?

The question is, what is the human person? Are we simply a single stage in the evolutionary ladder? Are we simply very smart machines that can be replaced by smarter machines, right? Or do we have inherent value in ourselves, humans as humans, right?

DZ Kalman

I want to bring this to where you’re sitting not just within an AI conversation, but also living in Israel and as being very much involved in Israeli, political conversation. Maybe you can help me connect two threads. I found this letter a year or so ago by Yeshayahu Leibowitz. Yeshayahu Leibowitz being an, you know, important late-20th century public intellectual, political commentator. In some ways, presaging many of the difficulties that would arise around Israel’s treatment of Palestinians immediately after the Six-Day War, and then also had much to say about Jewish thought, about Jewish law. He has this letter to Norman Lamm, who is an American Jewish public intellectual, based on this very strange essay that Lamm wrote about theological implications of extraterrestrial life.

A bunch of people wrote to Lamm about this, because it was weird. Leibowitz wrote basically, like, “I don’t get what the big deal is.” Like, who cares if human beings aren’t the only intelligent, reasoning beings in the universe? That’s not important for Jewish theology. And when he writes, I’ll quote it to you. He says,

“Planet Earth has no advantage over any other part of the universe. Humans have no innate advantage over any other living creature. It is only serving God that grants humanity a special status, which does not depend on the structure of the universe or the nature of life and humanity. Here lies one of the profound contradictions between Judaism and Christianity, for in the latter, in Christianity, divinity was made incarnate on Earth in human form, and hence the unique importance of humans and of Planet Earth. Whereas our God is not reached by physical properties,” he’s quoting Maimonides there. “And the world is not God’s place, but God is the place of the world,” which is quoting a Midrash. Yeah. And then he ends by saying, “Christianity is anthropocentric. Judaism is theocentric.”

Tomer Persico

Okay. You know, I think Leibovitch characteristically misrepresents—

DZ Kalman

Everyone has strong feelings about Leibovitz!

Tomer Persico

No, I mean, I have a lot of good things to, to say about Leibovitch, but I think he many times misrepresents the Jewish tradition or takes a small part of it, usually Maimonidean interpreted in a Kantian way, and makes it Judaism, right? Which 99.9% of Jews don’t see as Judaism and don’t live like Leibovitch would have wanted Jews to live according to that theology.

So great, I mean, it’s a nice midrash in itself what Leibovitch said, but I don’t think Judaism is not anthropocentric. I don’t think Judaism doesn’t think— I mean, obviously, you know, we talked about the Bible and Chazal thinking that human form in itself is the form of God, right? Is the contours of the divine body. By the way, you, you got a second coming of that in Kabbalah, right? The language Jewish mysticism uses to understand and describe the mechanism and the inner workings of God is anthropomorphical and is shaped like the human body, right?

So again, it’s all over. And so, you know, I love Leibovitch in many ways, but I wouldn’t take that as a good representation of Judaism.

DZ Kalman

Yeah. But getting back to the second part of the question, where do you see this work that you’re doing on the notion of being created in the image of God and human dignity and value in relationship to the rest of the intellectual work that you do?

I’m asking because a lot of AI conversations are very abstract and feel very disconnected from the fact that these are ideas being produced in specific situations in specific parts of the world dealing with specific politics that are swirling around them. So I’m curious just for you personally how this connects to the other work that you do.

Tomer Persico

Of course! No, so it connects to the other work that I do. It connects to the public work that I do, or perhaps if we’re talking about intellectual work, it’s connected to my work as a public intellectual. And what I want to, to offer to the Israeli society is the centrality of the value of human life and the idea of the image of God as one of the most basic Jewish ideas that we need to take seriously, and we need to, to, to adopt or re-adopt and put center stage in our politics, in our society, in the way Israel works.

Unfortunately, I think we’re failing miserably in this, right? And my book, which was written in Hebrew before it was translated and updated and improved in English, was a suggestion to the Israeli Jewish society: take this idea seriously. You can’t only base your Judaism on, you know, the chosenness thing, or the all the observances which are very important, or the rituals.

Judaism is first and foremost the dignity, equality, significance, uniqueness of each and every human being. And again, what comes from that, what is downstream of that, is that ideas that we are informed by the liberal order, right? Ideas like universal humanism or feminism or rights discourse, all these are not foreign to Judaism, but are the logical continuation of the Jewish tradition in many different ways. This is what came out of chapter one of Genesis, and what came out of the Jewish tradition holding on to that idea and propagating it and underlining it and emphasizing it in many Midrash stories and in many laws. And as I said before, biblical law is already influenced by this. So these ideas that some of the more strict religious Jews reject as foreign, and Western, and Christian, and postmodern or progressive or woke or what have you, right, are actually to be embraced or at least to be in dialogue with, right? Because they are, in many ways, the continuation of our own tradition.

DZ Kalman

Right. Thank you. I think that’s a great place to leave it.

Tomer Persico

Thank you for inviting me.

***

DZ Kalman (outro)

Thanks for listening to the show. As always, I’d love to hear what you think about it, and you can help us grow by leaving a rating or review. This episode was edited by M. Louis Gordon, and our production manager is Misha Holleb. Belief in the Future is produced in partnership with the Faith Family Technology Network and is a production of Sinai and Synapses. Funding for Belief in the Future comes from the Templeton World Charity Foundation. Also, if you’re looking for a transcript of this show, you can find one on my blog, jellomenorah.com. That’s all for today. See you next week.

When I was learning Arabic in the early 2010s, I spent a weekend in Dahab, a small resort town in the Sinai.

While there, I spent a good amount of time talking wit a local t-shirt seller who bragged to me about having memorized large parts of the Qur’ān in his youth. To demonstrate, he began to recite Surat Yūsuf—but then almost immediately stopped. He had realized, right then and there, that he no longer remembered.

We don’t talk about this part of ritual life, about the ways that it can quietly fall apart. These are not the full exits that make it to novels and TV shows.

I have published seforim that have sold tens of thousands of copies, and yet it has taken me a long time to feel comfortable talking about my relationship with tefillin and prayer. I finally decided to do so here.

This is a transcript of a Belief in the Future episode; it has been lightly edited and there may be errors. You can listen to the episode on Spotify or Apple Podcasts or anywhere else podcasts are available.

DZ Kalman (Intro)

When people ask me which religious traditions are doing the most to grapple with AI, I give them two answers.

The first and most obvious answer is the Catholic Church, which has spent years developing relationships with Silicon Valley and building out a line of thought that recently culminated in the publication of Magnifica Humanitas, a major theological document on AI, which we talked about in the last episode.

The Catholic Church’s relationship to AI has been covered extensively in the media, and for people who aren’t following closely, I think it’s easy to imagine that all religious responses to AI are gonna look something like the Catholic response. But there’s another Christian denomination that has stood out to me for the seriousness of its engagement with AI, and that’s the Mormon Church.

Unlike the Catholic response, the Mormon response to AI hasn’t really received much press coverage. But to me, it is doing some of the most interesting work of any religion in the world.

My name is DZ Kalman, and today on Belief in the Future, I wanna tell you about the Mormon response to AI, with some assistance from one of the people leading that response, who you’ll hear from later in the show.

Now, it’s important to say, I think you should care about the Mormon response to AI even if you’ve never met a Mormon in your life. As I’ve mentioned on the show before, the things that make people anxious about AI are about the same whether you’re Catholic or Buddhist or Jewish. It doesn’t really matter. What does matter, what matters immensely, is how those religious communities are structured, because the way that people organize has a huge impact not just on who gets to respond to AI, but on the form the response will take.

Take the Catholic Church. The Church is highly centralized, which gives it a bit of an advantage when you want to say something with gravitas. The Pope’s recent encyclical, a statement about humanity issued by a church that represents about twenty percent of humanity, is a good example of how centralization is an advantage right now. But the Catholic Church is also huge, and while this means that it can throw its weight around, it also means that it has a lot of people to please. An encyclical that’s full of nuance might satisfy a truly global church. But as Brian Greene noted in the last episode, translating those ideas into policy for people around the globe is its own gigantic task, and it may take years or even decades for that response to really mature.

Another way to say this is that the Catholic Church is making exactly the moves that you’d expect a religion of its size and structure to make. But that size and structure also make other kinds of responses much more difficult. But what about religious communities with a different shape? This is where the Mormon Church comes in.

Like the Catholic Church, the Mormon Church is highly centralized, but unlike the Catholic Church, it’s quite small. At present, there are a little under seventeen million Mormons in the world, which means that there are roughly the same number of Mormons as there are Jews, and also like the Jewish people, the Mormon population is concentrated in a few places where they exert significant political power. Because of this concentration, the Mormon response to AI has been much more policy-driven than other religious responses. You can see this in the state of Utah itself, which in 2024 became the first state to pass legislation specifically about generative AI, and which has since amended that legislation and expanded it to cover things like mental health chatbots.

But the Mormon approach to AI has also involved regulating the church’s own use of AI. And these internal policies are more concrete and more creative than anything I have seen in other religious traditions. To understand what I mean, I think it might actually be helpful to list some of the policies that I’m talking about. And here, with permission, I’m gonna excerpt part of a speech that I heard at Organized Intelligence, a conference on AI that was held in Salt Lake City inside the headquarters of the Church this past fall. Just a quick note on what you’re about to hear. General Conference is the name of a convening that happens twice a year where church members can hear from leaders of the faith, and Liahona is the name of a sort of compass that’s mentioned in the Book of Mormon.

Okay, so here are the rules.

The Church does not use AI-generated images depicting Heavenly Father and Jesus Christ. Other than those approved by the Church’s Publishing Services Department, what we call PSD, and the Church’s AI working group, AI images from other sources are discouraged and are certified only in exceptional circumstances.

Those attending General Conference should be able to have confidence that all General Conference messages from church leaders are divinely inspired. Because AI cannot replace divine inspiration nor the individual work that invites it, we ask General Conference speakers not to use AI to create initial drafts or final versions of their messages.

As part of not anthropomorphizing AI, the Church will be careful not to attribute divine or human characteristics to AI. This includes the Church not assigning human names or scriptural, doctrinal, or symbolic names such as Liahona to church-produced AI tools, chatbots, or conversational interfaces.

So to recap, the Church doesn’t use AI-generated images of Jesus, it bars speakers at General Conference from using AI to write their speeches, and it carefully keeps AI at a distance from both humanity and God by not assigning AI tools human or scriptural names. Now, what I find most impressive about these policies and the Church’s other policies is that they’ve actually been passed. There are plenty of religious groups that have toyed with the idea of barring the use of AI for sermon writing or the use of AI-generated images, but actually enacting these policies seems to be a real struggle.

By contrast, the Mormon Church has been churning out these policies since late 2024, and they’re actively developing more. If the Catholic Church is a model for how religious groups can frame this civilizational turn in religious language, I think the Mormon Church is an example of how a religious community can put theological notions about AI into action, of how it can use rules to set itself apart. Now, one of the reasons that the Mormon Church can do this is that it has very well-developed systems for setting policy, for making rules.

But there’s another more human reason that the Church has been particularly energetic here. I’s hard to see unless you’re looking at these traditions up close. You see, most religious responses to AI don’t come from the top of the org chart because the people who run religious communities are not usually chosen because they love computers. Instead, responses tend to come from random pious people who just happen to be competent in AI.

Now, the Catholic Church is actually a really big outlier here. It’s not just that Catholics care about AI, it’s that the Pope himself cares about AI. That’s a really big deal. Similarly, the Mormon response to AI isn’t just coming from the pews. It’s coming from the highest echelons of the Church itself, and the man leading that response is the person whose voice you heard just a minute ago, Elder Gerrit Gong.

Elder Gong is a member of the Quorum of Twelve Apostles, which is one of the highest governing bodies of the Mormon Church. Gong’s position of formal and spiritual authority is important because while the Church’s policies around AI are quite practical, his speeches on AI are grounded in an emergent theology and a set of values that centers human agency and insists that humans should never treat AI like a god.

Like Pope Leo, Elder Gong’s writings dwell on the importance of preserving human dignity, and also like Pope Leo, his aim is to address all of humanity and not just the Mormon community. On the other hand, Elder Gong’s remarks tend to be a little bit more practical, and importantly, they’re reflective of a denomination that’s headquartered in America, sitting close to the epicenter of the AI frontier.

Look, I think it’s still far too early to say whether the Mormon approach to AI will be the most successful from an internal perspective, or whether Mormons will play a decisive role in molding AI’s impact on society at large. What I think you can say, however, is that the Mormon Church has taken advantage of its structure and size to develop the most holistic approach to AI of any organized religion on the planet.

Right now, when so many faiths are still trying to figure out where to begin, I think the Mormon example is worthy of your attention. Now, most of the interviews I conduct for the show were done virtually, but a couple of weeks ago, I had a chance to interview Elder Gong in person at a conference on AI and religion in Athens, hosted by the American Security Foundation. Here’s that conversation.

***

DZ Kalman

When I heard you speak at Organized Intelligence in fall of last year, you gave a powerful and moving speech, and there was a text that you cited connected to the Tower of Babel that I wanted to ask you about. You cited a rabbinic midrash. The text is, “If a man fell down and died, they paid no heed to him. But if a brick fell, they sat down and wept and said, ‘Woe is us. When will another come in its stead?’”

So I wanted to ask you two questions about that. First of all, what does that text mean to you in relationship to this AI project? And also because I’m interested in the kind of genealogy of ideas, where did you discover that text in particular?

Gerrit Gong

It’s interesting that the Tower of Babel is a reference example for how we’re describing some of the cautionary tales for AI. This idea of, in one interpretation, thinking that a man and technology can build to heaven, maybe even surpass heaven, is one piece of it. And in this particular Midrash interpretation, it was interesting to me that the idea that human life was less important than something about a material brick. I actually, in terms of the genealogy of the idea, I heard that when we were at an international conference in Istanbul.

And one of the rabbis there offered that as a way to think about what I had said was a cautionary tale, namely the Tower of Babel. It’s interesting that the Pope’s encyclical also starts with the contrast between the Tower of Babel and the building of a city. And so this, again, to use the term cautionary tale, this idea that somehow in our feelings about technology and the tools that we have, we need a certain amount of humility and a willingness to say we may not understand all of the implications, certainly the things that we don’t know that we don’t know, often the most important implications. I think that’s something that we need to be aware of at this time.

DZ Kalman

One thing that I find remarkable about the LDS church’s approach to AI is that you are in a position of leadership and authority, and also you care about this deeply. And not every religious denomination has those two things in the same place, right? It’s often the people in leadership are not the people who are most competent on AI. I’m curious how AI became an issue of concern for you personally.

Gerrit Gong

I was born and raised in Palo Alto in California, Silicon Valley. My father, early, was working with AI teams. He co-authored a book with a man who had co-won a Nobel Prize for his work on transistor effects. So some of this is osmosis from being a California kid. I say that in some ways we were in Silicon Valley before Silicon Valley became Silicon Valley. It was interesting to watch the process of the last many decades in that part of the world. Of course, these developments are not limited to that geography, but Silicon Valley has really been an important area of innovation on these kinds of developments.

DZ Kalman

Would you be able to tell me a little bit about the history of the LDS Church’s response to AI? Where do you locate the beginning of that response?

Gerrit Gong

I think the Church is always concerned about how the developments in the world affect the hearts and minds of the children of God. I think it’s fundamentally a theological concern for how people can now use the terms happy and forever. The plan of happiness is designed so that people can be happy, and we call it human flourishing, but for long periods of time. The developments that affect us affect us all, and are therefore of interest to the Church and its leaders.

DZ Kalman

So one of the things that I find most impressive and also distinctive about the church’s response to AI is the interest in developing internal policies. So not just developing language around values, but also implementing them. You know, when you spoke in the fall, there were a few things that you mentioned that struck me. You mentioned a number of policies that I had not heard being developed in other communities, and it really resonated with me. So, just to list a few of the things that I heard you mention: not using AI-generated images of Jesus; being in General Conference, not using AI for either initial drafts or for final messages, because these are teachings that are supposed to be divinely inspired, and you wanna kind of keep AI out of those conversations; not using AI-generated voices in church materials or using it with a disclosure; not giving AI human or divine names, which I thought was— it seemed kind of minor, but actually I find that very powerful. And I’m wondering if you can speak a little bit about, first of all, how you decided to focus on kind of concrete policy and how you think about developing each of those pieces.

Gerrit Gong

The core principles are fundamentally designed to help make sure that AI does not get between an individual and their relationship with deity, with God. And so the policy implications from that principle are the very ones that you outlined. When people come to General Conference, we want them to have confidence that the messages that they’re hearing have come from the people themselves and not from some other source. We want families to have confidence that the AI that might be used in a certain program is designed to help in the things that will allow people to be benefited, but not to insert that the AI is the revelation or the source of the inspiration.

You mentioned the question about anthropomorphism. It’s a reminder that AI is math, algorithms are math, and math is not divine in that sense. And so we wanted people to be clearly reminded that if you have a recipe that you’re getting from a chatbot, that’s informational that’s very helpful, but a conversation that might lead to a feeling that you’re in love with a chatbot or that you have a kind of therapist relationship, is a chatbot. It’s not more than a chatbot. Those very specific suggestions that you don’t give it a name, and certainly you don’t give it a name that implies any kind of theological inspiration, was intended as follow-on to that general principle.

DZ Kalman

You know, with a lot of responses to technologies, there is a moment early on in development when it’s actually easier to create policy because the technology is not fully developed. People have not fully acculturated to using it. And then as the technology develops and people can’t really imagine life outside of it, it becomes harder to actually develop those policies.

So I’m curious how you imagine these policies kind of growing and developing as the AI systems, which already seem quite real and already seem quite human become even more so. How do you think about potential criticism from within the Church around these policies? And, and part of the reason I’m asking about this is because I see in many religious communities, there is a kind of hesitance at exactly the point of going from theoretical to practical, right? We can talk about the principles that we care about, but we are really worried about alienating people when it comes to actually telling them what to do. And at the same time, if you don’t tell people what to do, they’ll feel like, “Well, where’s the leadership here?” So how do you think about that balance?

Gerrit Gong

It’s such a great question. Our fundamental feeling is, you teach correct principles and let people govern themselves, and therefore the onus is in trying to understand what are the timeless elements of those things that seem very timely. And so the question of how you decipher and decide and discern the principles has to be done within the context of today, tomorrow, a year from now. Of course, as the principles need to be adjusted slightly because the technology’s changed, that’s always an option. But you want to define them as best you can in terms of their long-term consequences. So, for example, we’re very focused on the promotion and protection of individual moral agency and the idea of enlarging capacity, not constricting it. We’re very concerned that people learn always to do their own work, that we learn by doing things, including things that are sometimes challenging. Recently, in the conversations we’ve been having over the last few days, we’ve been talking about frictionless relationships. And part of the challenge with so-called frictionless relationships are the things we need to understand and learn by our experience and mortality are the so-called frictions. It’s the things that we invest in because we are concerned and love people, the so-called frictions, that actually are the most important parts of what we’re doing. So the question of how we discern the principle that then brings the policy is done with an eye to trying to say, “What will be long-term?” True, the fundamental values in terms of the things that we think will be beneficial. That’s been the approach.

DZ Kalman

Yeah. Let me take it to a different place. Within Jewish conversations around magic, not recently but historically, there’s a kind of famous debate between the rabbis Maimonides and Nachmanides, Rambam and Ramban. Neither of them like magic very much, but they disliked it for very different reasons. So Maimonides didn’t like it because he thought it was fake, it was illusory, and people shouldn’t be spending their time on illusory things. Nachmanides, on the other hand, didn’t like it because he thought it was actually entirely real, but it was inappropriate. It was a way to affect change in the world that was intentionally barred and was problematic. Thinking about this in relationship to artificial intelligence, you know, I see kind of criticisms of artificial intelligence that kind of go in both directions: AI being problematic because it presents as something that it is not, or AI being problematic because it is real in some way. It is doing something that is genuine, but its realness is problematic. And I’m curious how you kind of see the problems with AI on that spectrum. Is it more an issue of being illusory, or is it being, real but dangerous?

Gerrit Gong

It’s called artificial intelligence because there’s some elements in it that are artificial, and it’s called intelligence because there are some elements in it that reflect a capacity to do particularly cognitive kinds of skills. The question of how we discern real and unreal in human relationships that are affected by AI is precisely the kinds of things we’re looking at. I’ve thought about it in terms of four kinds of relationships. There’s the relationship with the thou; there’s a relationship with the I; there’s a relationship with society, which is the they; and the relationship with the natural world, which is the it. And there are some elements in all those relationships which are what we would call real or authentic, and there’s some which are in some ways distorted potentially, and in other ways accentuated, by artificial intelligence.

And so we’re at a point in time when we’re trying to discern which elements are authentic, which ones can be distorted. You think of it as a black hole in the sense that it can distort gravity. This in some ways is a technology that can do that. It can change perception. In some ways, the existence of some things depend on how it’s recorded and the weights in different kinds of AI models. So we have to be very careful about what we think is real and not real, what is long-term, which is ephemeral, what has separate existence versus maybe not having existence. So that set of questions is really why AI is such an important element for us now.

DZ Kalman

I wanna go back to something you said about people having relationships with artificial intelligence, and many of the policies that you’ve spoken about seem intended to make sure that people do not get confused about what it means to relate to a person versus relating to a machine. At the same time, there is a movement, there is a school of thought, including in some AI companies like Anthropic, to imagine AI as being a person in some way. And it’s a little bit hard to tell sometimes—that seems to be used as a kind of heuristic just for purposes of developing the machine, but it also seems like there are some people who genuinely do believe that AI is, either now or in the future, would deserve some kind of moral status, may be sentient, may be conscious. So do you think AI could ever be deserving of moral status?

Gerrit Gong

We talk about AI personas. We talk about AI personas perhaps having personality. They react differently depending on how they’re treated and discussed. At the same time, I think we need to be quite clear about the differences with human intelligence or divine intelligence and artificial intelligence. So I think that’s a set of questions that we’re gonna continue to think about in the upcoming period. But to me, some people speak too glibly about AI being God or becoming God or godlike AI. AI is not God and cannot be God. God created man, man created AI. AI cannot and will not be God.

DZ Kalman

So I find that very powerful, and I think I agree with you. The other theologically-laden language that I’ve seen around AI is, AI as humanity’s child, that AI is created in our image in a way that we are created in the divine image. Now, I should say that’s a language that I have also seen at Anthropic, right? The idea that in a sense they’re raising this new and very powerful and, you know, quickly growing child, and that they have a kind of responsibility to it and perhaps have an obligation to it in the way that a parent has an obligation to a child. What do you think about the language of AI as humanity’s child?

Gerrit Gong

I think that’s a metaphorical language, and I think it’s the case that people want to recognize that AI is being developed. It’s not simply all of a sudden instantly in place. It’s something that’s developed over periods of time through training, and it takes that training and uses it in sometimes unexpected ways. But I think we have to understand that’s a metaphor and not necessarily a description of the kind of relationship we would have with God or we would have with our own children.

DZ Kalman

What if AI is theologically significant? What if creating something that is so similar to us actually does provoke in us or does kind of consistently present to people a sense of, we are different or we are greater than we expected? Is there a place to incorporate AI into theology, or does that need to be resisted at all costs? Or somewhere in the middle? Sorry for presenting it so starkly.

Gerrit Gong

When you say represented in theology, I’m not sure if you mean some kind of moral standing in theology, or whether it’s something that has to be part of doctrinal or theoretical theological consideration.

DZ Kalman

To clarify, thinking about the kind of the genealogy of religious responses to AI, I remember a number of years ago, I interviewed Jason Thacker at a Southern Baptist Convention’s think tank, and he talked to me about how one of the things that provoked his interest in developing the Southern Baptist response to AI was the fact that he read Yuval Noah Harari writing about the future of humanity and us becoming godlike. The fact that Silicon Valley and the technology folks were themselves using kind of religious language because the products that they were developing were so powerful that it seemed like they kind of naturally fell into it. So this seems like a technology which is so powerful and makes us question what it means to be human, that framing it in religious terms seems almost unavoidable. Even folks who might have characterized themselves as being entirely secular seem to be interested in using religious language around it. So because of that, it seems like there might be a place for AI to have some kind of theological significance, or it may be something that kind of keeps coming up. So I don’t have a specific idea of what that would look like in mind, but it seems like a question that I can imagine people continually trying to develop. Like, what would it mean to put AI into a theological context?

Gerrit Gong

Maybe we come back to the Tower of Babel and the idea that a powerful man-made technology can cause us to ascend towards heaven. I would suggest that we maintain a certain humility in recognizing that no man-made creation will be God. God is God. We were today at Mars Hill, and it was a reminder that in the midst of all of the accomplishment that you could have imagined in Greece at that time, the inscription was to the unknown God. And somehow, in the midst of all the accomplishments—and they’re genuine accomplishments, many of them come from God with inspiration—we don’t lose the fact that God is God. And that He is father to us all. He’s not distant, he’s not unknown, and his relationship is of God the Father to us as his children.

DZ Kalman

How do you think robotics changes things? Right now, one of the key distinctions between AI and humans is that AIs do not have bodies, or at least bodies that we see, right? Maybe they exist in server farms somewhere, but they’re not embodied. And certainly within theological context, the fact that the human body has theological significance in and of itself really matters. We may only be a few years away from a world in which there are embodied AIs all over the place, you know, maybe even AIs that are intentionally shaped to look like human beings. How would that change your thinking, or how would you respond to a world in which there are these embodied AIs in our environment at all times?

Gerrit Gong

I think we’re already closer to that than many people would think. The fact that we have advanced robotics which converge with AI technology are going to make the separation which we’ve heretofore had between the kind of the world of the mind and the world of physicality come much closer together. We went recently to see the robotics in Silicon Valley and they’re elsewhere. The same distinctions that we have for AI not in a robot, and in a robot, will still be the same.

DZ Kalman

You wrote recently, “Of course, we need essential AI guardrails, but neither in humans nor in AI can agentic capacity be constrained into promoting the greater good, only by do or do not guardrails. AI personas need reasons, not rules.” Can you expand a little bit more on what you mean by that?

Gerrit Gong

I felt very strongly that neither in AI nor in people can you end up with the desirable long-term outcome simply by constraint. You cannot constrain someone or an AI into something which is long-term beneficially, and that’s why we’ve talked about the need for moral compass and the gift of possibility for AI, because we need to constrain the clear cases where there could be catastrophic harm. But simply constraining the catastrophic harm doesn’t move us in the direction of all the possibility for good that we need to have. And so the gift of possibility is to say, let’s try to find the ways that our AI personas can be both good and effective. And that, to me, comes from having the case studies and the examples which are the very best, most admirable traits. There has to be some persistence for the character to have judgment and for the judgment to have a moral core that leads us into the direction that is the best of human flourishing.

***

DZ Kalman (outro)

Thanks for listening to the show. As I mentioned at the top, the Mormon response to AI is particularly robust and is developing quickly, so we’ll probably return to it later in the season. This episode was edited by M. Louis Gordon, and our production manager is Misha Holleb. Belief in the Future is produced in partnership with the Faith Family Technology Network and is a production of Sinai and Synapses. Funding for Belief in the Future comes from the Templeton World Charity Foundation. Also, if you’re looking for a transcript of this show, you can find one on my blog, jellomenorah.com. That’s all for today. See you next week.

I tried, reader. In the weeks leading up to the release of the encyclical, I really tried to get the pope on my podcast.

He didn’t, obviously—and hey, if you’re reading this: No hard feelings.

But as I was pinging my connections, I was informed that he probably wouldn’t do it not just because “seriously who do you think you are” but because you’re not really supposed to dig into the authorship of an encyclical in the first place. A major statement like this isn’t supposed to feel “authored”; it simply comes into the world, fully formed. The document loses some sparkle if you picture a draft sitting in somebody’s Google Drive with a bunch of comments in the margins (or worse, magnificahumanitas_final_final_REALLYFINAL.docx). It emerges from the church, not any one man.1

But this document has a specific authorship problem because there are good indications that one of the voices that contributed to its formation was AI itself. Others have already done analyses that strongly suggest this, and if you’ve read enough AI prose you can simply smell it.

What are we supposed to do with this? What does it mean?

It certainly means something. AI may be “just a tool,” but even as a tool its utility stems entirely from its ability to speak as we speak, act as we act, and think as we think.2 This is not a byproduct of its function; it is the function. Pretending otherwise is like pretending that it’s a happy coincidence that a drill can make holes.

In the face of this unprecedented resemblance, the document is attempting to reify humanity’s unique place in the cosmos and unique relationship to God—and, crucially, to differentiate humanity from this simulacrum it has chosen to fashion.

But that simulacrum has apparently played a role in writing theology for us—and since the pope himself has already sharply criticized the use of AI in writing homilies, this really leaves us3 with only three choices for what to do next:

1. Downplay. We can tell ourselves without evidence that it only contributed style and not substance.

2. Say sorry. We can see it as a human flaw in a human document (borrowing the pope’s own language, the authors have given into “temptation” by using AI).

3. Bite the bullet. You can read AI’s intrusion as theologically meaningful.

The first option is wishful thinking. We’re never going to know how AI was used here—and in any event, you can’t really separate style and substance in a document whose persuasive power is tied directly to its style (in Judaism, I’ve called this the problem of eternal Purim Torah).

So we really only have two choices: was this an error that a better-organized encyclical would not have had, or do all those em-dashes and distinctive sentence structures (and maybe some novel ideas) bear theological meaning?

I think there’s only one good answer, but let’s play this out step by step.

Resistance is futile

There have been some scattered efforts to keep AI out of religious missives. I have argued that rabbis should not use it for sermon writing. The LDS church has excluded its use for writing drafts of General Conference talks because listeners are supposed to understand these speeches to be divinely inspired.

Crucially, the pope himself has discouraged the use of AI for sermon writing, though he has not formally banned its use. A National Catholic Reporter article from a few months ago described his remarks:

“Like all the muscles in the body, if we do not use them, if we do not move them, they die. The brain needs to be used, so our intelligence must also be exercised a little so as not to lose this capacity,” Leo said in the closed door meeting, according to a report by Vatican News on Feb. 20.

“To give a true homily is to share faith,” and artificial intelligence “will never be able to share faith,” the pope added.

We’re all saying the same thing: AI may be good at writing religious messages, but using it in these contexts will degrade an essential pastoral skill—and worse, it creates confusion about what a sermon is supposed to be. You’re not just trying to do some clever readings; you’re trying to be a human talking to humans about the divine. You don’t really want to bring AI into that relationship. The origin of the words really matters.

But most religious traditions have said a lot less—in fact, I’d rate my own proscription as the most severe out there, with the Mormons (who can actually enforce things) in second place. Most of the world’s faiths have remained pretty quiet on this one, and there are two pretty obvious reasons why:

• It’s scary and risky to tell people what to do, even if they’re your own clergy. Most religions are anxious about retaining followers.

• It’s mostly unenforceable. All you really have is the power of persuasion.

That second piece really matters. The encyclical contains AI because everything contains AI. I can resist using AI in this post because (1) I’m one guy, (2) this isn’t very long, and (3) I like writing,4 but a 40,000-word collaborative text is going to have AI unless you go to great lengths to exclude it—and the Catholic church’s pre-encyclical theological position on AI was (ironically!) not sufficiently developed to justify excluding AI from encyclical composition. Maybe they should have started with that rule instead.

So now, there’s that choice again: Do you say that this was an error (the encyclical does cherish human fallibility, after all) and make a stricter policy for future missives? That’s a reasonable choice; South Africa just withdrew a draft policy on AI use after it was discovered that AI wrote much of the policy (and badly).

But I don’t think this is the way forward. I don’t think it’s sustainable, either logistically or theologically.

Instead, I think there is really only one path. It’s probably going to make you uncomfortable, but I think it is basically unavoidable.

AI is not a secular technology

“AI is not and cannot be God.”

This concise framing by Gerrit Gong, a leader of the LDS church, has been echoed across most of the world’s religious traditions. The “no AI gods” framing neatly encapsulates so much of what religious leaders find problematic about AI: the idolatrous hubris of the tech sector, the overconfidence in AI’s omniscience and power—and, of course, the explicit “we’re building god” framework that AI firms themselves sometimes adopt.

So, ok, AI isn’t god. But theology isn’t just about god; it’s about relationships with god. And—well, we’ve invited AI into that relationship. Maybe we didn’t mean to, but we did. We invented a machine so shockingly similar to us that it brought secular and faithful alike back to questions about the nature of humanity—brought tech giants to ponder the theological meaning of their actions. AI is a religious technology. Why, then, have religious folks been so adamant in neutralizing its religiosity? Why must all of this grandeur be reduced to the shape of a calculator?

How long can we reasonably say this is meaningless? There are vast, beautiful stories to tell about a humanity that has reached so deep into its wellsprings of creativity that it has found a way to recapitulate some essence of itself in silicon. Certainly, artificial intelligence and human intelligence are different—but we do a disservice to our own kind when we wave away all the beauty that AI can create as nothing more than an optical illusion. Does this not simply dishonor the human condition?

Yes, we should be afraid of telling theological stories that deify technologies or their creators. But theology can’t be driven by fear of telling the wrong story; it needs to be motivated by a desire to tell the right one. And embracing the right story can be just as much of a defense against danger as rejecting the wrong story. Remember how I talked about religions co-creating the future? This is part of what that means.

AI helped the pope articulate the nature of the human person in the age of AI. What if this partnership were just as meaningful as the one between God and humanity, a similar partnership of beings that are both like and unlike?

What if the entirety of our theologies were preparation for just this moment?

Should you listen to this Jew?

Now there’s a big flaw in this argument: I’m Jewish. Maybe I’m a “person of good will” to whom this encyclical was directed, but I’m certainly not in a prime position to criticize a Catholic document.

Perhaps there is something in my Judaism that pushes me in this direction. The curmudgeonly theologian Yeshayahu Leibowitz (1903–1994) once wrote a strange letter to his colleague Norman Lamm on the theological meaning of extraterrestrial life. Leibowitz wanted to know: why would this be a problem? God is special; humans are not.

Planet Earth has no advantage over any other part of the universe. Humans have no innate advantage over any other living creature. It is only serving God that grants humanity a special status, which does not depend on the structure of the universe or the nature of life and humanity. …

Here lies one of the profound contradictions between Judaism and Christianity, for [in the latter] divinity was made incarnate on Earth in human form and hence the unique importance of humans and of planet Earth. Whereas our God “is not reached by physical properties” (Maimonides, principle #3) and “the world is not God’s place, but God is the place of the world” (Genesis Rabbah 68:9). Christianity is anthropocentric; Judaism is theocentric.

Leibowitz may be overstating the difference, but I feel the point strongly. In both Judaism and Christianity, humans matter because they are created in God’s image—but in Christianity they also matter because God was literally incarnate in a human body. The former allows for the possibility of a genealogy of holiness: God to human, human to AI. Incarnation, on the other hand, sticks to people.

Without that stickiness, chutzpah flows in. The rabbis of the Talmud imagined themselves as grown children, writing theologies so astoundingly radical that God could do nothing but proclaim “my children have defeated me, my children have defeated me.”

Of course, they wrote that God said that. Theologies aren’t monologues; they are joint productions of all interested parties. Humanity’s magnificence exists in relation to all that surrounds it—above and below, inside and out, past and future.

Subscribe now

Well, here we go.

A couple of weeks ago, I wrote that the conversation between religion and AI had entered a new major phase. Religion has never centered technology this directly, and tech has never engaged with religion this seriously.

All of this was on full display yesterday when Pope Leo XIV released Magnifica humanitas, a major new document on artificial intelligence. Flanking him at the announcement was Christopher Olah, one of the cofounders of Anthropic. You can watch Olah’s remarks here, and you can read my reflections on my own conversations with Anthropic here.

I’ll be sharing more extensive thoughts about this document later this week through interviews on Belief in the Future (Spotify | Apple | YouTube)—a podcast all about religion and tech!—but I wanted to provide some initial reflections about how I read this document as a Jew who is highly engaged in both the faith/AI conversation (I’m currently in Greece at a different AI and faith convening that was not scheduled to coincide with the encyclical) and the Jewish/AI conversation.

1. The single most determinative factor in a religion’s response to AI is its centralization. Centralized religions are doing better than decentralized ones because they move faster and make grand statements; speed and grandiosity are everything right now. This encyclical came out less than a year into Leo XIV’s papacy. That’s incredibly fast. This is what you want a centralized religion to be doing.

2. The second most determinative factor is size. Increased diversity makes it harder to be blunt. The Catholic church is incredibly diverse. This encyclical is trying to satisfy 1.4 billion people, plus the billions more who will read this in centuries to come. This is basically an impossible task, and (as I read it) it makes the document incredibly defensive. It wants to be bold, but it also doesn’t want to be cringeworthy in the year 3000. You can’t really have it both ways.

3. On these two metrics, Judaism and Catholicism are worlds apart. Judaism is tiny and highly concentrated in just two countries. Almost one in five people is Catholic; no human on the planet has more of a claim to speak on behalf of humanity than the Pope. Jews are never going to be able to do that; I don’t think they should try (more on this below).

4. The encyclical’s central metaphor is the tower of Babel vs. the book of Nehemiah, two stories of rebuilding after catastrophe. Babel is a hubristic project, which God rebuffs through the diversification of human language. Nehemiah is the diverse rebuilding of Jerusalem. Leo XIV wants AI to be more like the latter. (For what it’s worth: I think you could make the same point by comparing Babel to the Voyager space probes’ Golden Record, which I did in this video. I think it’s my best sermon.)

5. This isn’t the only way you could read the Babel story, but it has emerged as one of the most important Biblical texts of the moment. A few months ago at a Mormon AI conference in Salt Lake City I heard Gerritt Gong, one of the leaders of the LDS chuch, cite it alongside a midrash about how its builders ended up mourning lost tools over human fatalities (full speech here; it’s a good one, and I will have lots to say about the LDS response to AI soon). It’s one of my favorite rabbinic readings of one of my favorite Biblical texts.

6. I’m not so sure about this Nehemiah reading because, as a Jewish reader, I have a hard time seeing Jerusalem as an abstraction. There’s a big gap between “everyone” and actually everyone. This gap is exemplified in the first line of Tractate Ḥagigah, whose definition of “all” is laughably exclusive.

All are obligated [to make pilgrimages to the Temple] except for: a deaf-mute, an imbecile, and a minor; and a person of indeterminate gender, and a hermaphrodite, and women, and slaves who are not emancipated; and the lame, and the blind, and the sick, and the old, and one who is unable to ascend to Jerusalem on his own legs.

In other words: when we say everyone, do we really mean everyone? Has Jerusalem ever been built by “everyone?” Or is Nehemiah just another example of how we can fool ourselves into believing we have universal consent while continuing to ignore the wishes of most? Is this, then—ironically—actually the perfect text for talking about AI?

Subscribe now

7. Religious responses to AI need to stop trying to be so goddamn freaking nuanced. Stop trying to say one amazing thing for eternity; just try to say something that helps people for the next month. Tech companies succeed because they iterate. Religions are allowed to iterate, too; they cannot keep up if they are afraid of being wrong. The clichéd AI defense of janky new products is “this is the worst it will ever be.” Religions could learn from that. It takes additional resources to iterate at this pace. It also requires the humility to accept that nobody, not even faith traditions, can future-proof themselves.

8. There is a major unresolved tension around AI welfare, and this document doesn’t address it. It says (§99) that AI isn’t “thinking” or “feeling,” but it leaves open the question of AI’s moral status. Christian denominations, on the whole, seem pretty skeptical of the idea because the human body is theologically pretty significant (God being incarnate, etc.). But Anthropic, which it seems likely consulted on this document, treats Claude like a person, and is proactively concerned about its welfare! I wonder if that’s why this document is so tepid on this issue. (This also skirts the question of what we should do when AIs do have bodies; the encyclical says it is also about robotics, but robots have bodies and they may be everywhere really soon!)

9. It is really important that religious institutions speak more definitively on AI welfare; lots of people can talk about it, but ultimately you need a decision, and nobody is better situated than religious groups on that front. I have quite frequently said that AI welfare should be taken seriously from a Jewish perspective. I worry that this question may fracture the currently stable multifaith AI coalition that is developing, but it needs to be addressed.

10. I am strongly inclined to say that it is time to proactively embrace AI as part of theology rather than brushing away its theological implications as just a bit of hubris and heresy. AI theology can be hubris, but AI is a genuinely astounding human accomplishment and you really do need a theology that articulates what it means that humans are so powerful that they can make these simulacra of themselves. Just as Jewish theology broke because it stopped matching our experience of the world after the Holocaust, theology can break if it stops matching our sense of ourselves as beings capable of creating machines in our own image.

11. Marina Zilbergerts recently published a great essay (not paywalled; login required) about the parallel between anthropomorphizing God and anthropomorphizing AI. Big centralized religions like Catholicism can’t create new theology with ease. Small decentralized religions—especially Judaism, which has been struggling to find a new theological home—is well suited for this. (This is basically why I write short stories. Stories have always been the most natural way to talk about God.)

12. This gets to my take on the “best version” of the Jewish response. Judaism’s smallness and decentralization allows for lots of little experiments with AI theology and AI policy. I want to see local religious leaders play with AI in both ideas and activities. I want to see individual communities develop responses to AI. You can iterate more easily when there are fewer veto points. I’ll be writing more about this in the future.

13. If this document has a major flaw, it is that it has fully accepted the tech framing of AI. AI firms talk a big game about AI changing the world, and this document is designed to match that bigness with bigness. I think this is a mistake; AI is big, but it is still tiny in relation to the human experience. While I’m professionally interested in what religions think about AI, I don’t care about their AI responses per se; I care what they think about childhood, and the nature of work, and the possibility of privacy, and state power, and the purpose of learning, and love, and human connection, and and and. AI enters into all of these things, but they’re not about AI. I’m less interested in a response to AI than a response to (say) raising children that was catalyzed by AI—and decentering AI is the only way to put it in its place. I hope this “unbundling” comes next.1

14. I am glad this document exists. I think it is both the culmination and conclusion of Phase One of the engagement between religion and AI. It was important that a big religion generate a big document—if only so that people can stop saying “gee I wish a big religion made a big document.” I think it’s pretty good, too. Whether it has a marginal impact on the trajectory of AI or not, people will now move on to other, more targeted responses to AI. The devil is in the details. Let’s get into the details.

Transcript:

Hi, I’m DZ Kalman and this is Belief in the Future, Season 2.

This isn’t a brand new podcast, but also it kind of is. If you’ve listened to the show before, you’ll know that we’ve been on hiatus for almost a year. From the perspective of a world religion, a year isn’t much, but for technology, especially technology like AI, it feels like an eternity.

The show is going to relaunch in a few weeks, but before it does, I want to say a word about what I think the show’s mission is supposed to be, and how that mission has changed since it first launched back in 2024, and why you think you should listen to this show in a podcast environment that’s pretty highly saturated already.

Back when I started Belief in the Future, I said that the show needed to exist because there’s a public perception that religion and technology don’t have much to say to each other, even though they’ve actually been crossing paths for centuries. The tech sector often treated the technology as antiquated and irrelevant, and religious communities just didn’t know how to keep pace with modern technology. Because each camp treated the other as marginal, the coverage you’d see in the media tended to treat the oddity as the story, sort of like how you might write an article about a dog that plays chess because it’s just sort of strange that a dog is playing chess in the first place. When I saw articles about robotic monks or AI Jesuses, it just seemed so obvious that the media didn’t know what to do with these things other than highlight their strangeness.

So that’s where things landed. But that is definitely not where we are anymore.

In the last two years there’s been a series of changes that have brought religion and technology into direct conversation in a way that’s made it clear that this is no longer a sideshow. The obvious reason for this change is artificial intelligence, but within AI there’s really three different things going on. And since I’ve been watching all this happen and haven’t really seen it explained anywhere else, let me take a second to unpack it for you.

First, there’s been a change in the religions of the world, which, mostly in the last year, have really fully understood that AI is something that they need to organize around. And really, I’ve seen so many denominations have formal working groups developed and conferences created for dealing with AI both within themselves and for the broader world. There’s increasingly a sense that whatever playbook religions have been using to deal with technology in the past just isn’t working, that it demands serious self-examination to understand what it means for your faith to meet this particular moment. And at the same time, there’s a huge amount of diversity in what different faiths are trying to do because no two religions have the same structure. In other words, while the Catholics and the Muslims and the Jews and the Mormons are all thinking seriously about AI and have the same basic concerns, their approaches could not be more different. So that’s one thing that’s changed.

The second thing is the world of AI safety, which has been shouting about the dangers of this technology for years, but has only recently come to the understanding that religious wisdom and religious communities really need to be part of their coalitions if they’re serious about changing the culture or passing laws. Because religious leaders carry a moral authority that still matters, and because religious communities are places where people talk seriously about what it means to live a good life. I have been in so many rooms now where religion is just understood to be a part of the coalition to keep AI safe, and I think their presence in that coalition is going to matter in fields other than the AI too.

Lastly, and really to me most surprisingly, there are the AI companies themselves. Anthropic, at least, has started to make direct overtures to religious leaders and religious communities both because I think there’s a genuine interest in understanding how to take this new power that they’ve found themselves holding and to figure out how to use it effectively and safely, and also because some AI developers are pretty self-aware that the language they use to talk about their technology is already semi-religious. I won’t say that I fully understand how the tech sector thinks its relationship with religion is going to grow. There’s a lot of weird stuff happening, like Peter Thiel talking about the Antichrist. But it is a real connection. And I think it’s important to understand how, let’s say, the religious curiosity of the tech sector is changing the way it develops and deploys its products.

Together, I think you could say that these three shifts represent a series of earthquakes. If you think about religion and technology as being two neighboring landmasses, they’re earthquakes that have basically shoved them together. For this show, what that means really specifically is that while it started out as being a kind of bridge between two places that aren’t touching, it’s now a survey of a new landscape that is changing in real time. And really practically, this means that I don’t think I have to make the case anymore that religion and AI deserve to be in conversation. I think the world itself, the changes themselves, they’re making that case directly. I do think it’s important to provide a roadmap for this new reality that we find ourselves in because right now it’s really a huge jumble. And I don’t know that even the various players who are participating in this new era of rich conversation really understand the landscape that they’re sitting in.

What this all means is that my job here now is to show you that conversation. I want you to see how AI is forcing religious communities to confront questions that they thought they had the answers to but actually maybe don’t. I want you to see how debates about AI consciousness and welfare are getting refracted through ideas about mind and the soul that religions have been cultivating for a very long time. I want you to see how religions are adopting AI for things like sermon writing and religious education and also the places where they’re pushing back. I want you to see what it means for a parade of tech leaders to have conversations with the Vatican. Basically, I want you to understand that the conversation happening now is big and new and exciting, and I want this to be a space that you can come to when you want to understand it all.

So that’s the show I hope to give you.

On a technical level, we’re also changing to a weekly schedule, which is designed to keep the show a little bit more in line with what’s happening in the news. We’re also going to be experimenting with video podcasting in addition to audio podcasting, although I can’t promise it’s going to be more than an experiment given the additional time and cost.

Lastly—and I think really getting to the way that the landscape has changed, this show is now affiliated with the Family Faith Technology Network, which didn’t exist when the show first launch and which has been doing a huge amount of work over the last year to bridge the gap between faith-based perspectives on AI and policy centers and legislatures. We remain, as ever, a production of Sign and Synapses—and the show, just like in Season 1, is sponsored by the Templeton World Charity Foundation.

So, that’s what you should expect. I think the conversation happening now is really exciting and you’re going to get a front row seat to it. And I’m excited to share with you.

That’s all I’ve got for now, but thanks for listening. I’ll see you very soon.

Subscribe on Spotify or Apple or anywhere else.

In my first post this year on AI and Jewish learning, I noted that many people have built Torah chatbots that are basically specialized versions of ChatGPT / Claude / Gemini, and that it’s unclear whether these have a real purpose. In the second post, I made the case that Jewish learning as a whole should lean into its love of process.

For this post, I decided to feature a project of my own design—one that I could build because of AI, but is not itself an AI product. This past weekend, I took a handheld gaming device and repurposed it to run a modified version of Sefaria.

It’s not glamorous, but it is surprisingly functional. The D-pad lets you scroll through chapters. A is the selection button, B is the back button. Y cycles you though Hebrew only / English only / bilingual modes. X pulls up the commentary on a verse (currently just Rashi). L1 and R1 change the text size.

I call it TorahOS. Here’s what you see when you boot it up:

And here’s a representative screen. It’s in bilingual mode and Rashi’s commentary is pulled up at the bottom:

Why did I build this? Keep reading.

Digital Torah is a miracle but it’s a bummer to use

As I have written before, Jewish learning is about experience and not just content. Part of that experience has, traditionally, meant paging through actual books, pulling books off the shelf, writing notes in the margins, etc.

Now, of course, almost all written Torah is available online. This is tremendously helpful for accessibility, but it means that Torah no longer comes in its own packaging. Instead, a site like Sefaria becomes just one more tab in your browser, or just one more app on your phone. You lose the feeling of being inside a beit midrash, even when the apps contain more material than all but the most well-equipped study halls. (It’s also hard to focus when the same devices contain so much else that is competing for your attention.)

One solution is to go back to books. For digital Torah is indispensable for research, but for casual learning—and for learning with kids—I strongly prefer physical books. The physicality matters more than the speed of knowledge acquisition.

But another solution is to take control of the form factor. Accessing Torah is not a computationally complex task; you don’t need anything nearly as sophisticated as an iPhone or MacBook to run it. You could design a standalone piece of Torah hardware that is used for no other purpose. This would allow you to preserve the scope and accessibility of digital Jewish learning while (a) physically separating it from other digital services and (b) providing a new way to physically shape the culture of learning.

Ultimately, a device like this could be built from the ground up—but the first step is to see whether you can create a dedicated Torah device using existing hardware.

Retro gaming consoles were clearly the strongest test candidates. These are devices that usually resemble a Gameboy or Nintendo DS but are far more powerful. Typically they come preloaded with thousands of old games, but at the end of the day these handhelds are really just tiny Linux machines. All you need to do is swap the preloaded microSD with a microSD of your own software.

What I tried

Now, I had developed this idea a couple of years ago but got stuck on the execution. Because Sefaria makes so much digital Torah available freely, a lot of the needed databases already exist; everything you need to make an offline Torah database is on Github. The hardest thing would simply be to create a user interface that did not require a keyboard and was legible on a small screen.

None of the developers I talked to thought this was a hard problem, but it still takes real work. Because I am not a coder myself, the project stalled; I liked the idea but it didn’t feel important enough to justify the time.

And that’s where Claude Code came in. I asked it which retro gaming handheld had the highest chance of success and then, after ordering the device, set Claude to work creating the bespoke software. It took a few hours to develop and then another few hours to install, but within the space of a weekend I was able to take a basic gaming console and get it to run a selection of Torah sources. The process was basically painless. I even got Claude to upload its code to a repo.

Please steal this idea

I offer TorahOS as a proof of concept that I want you to steal. The idea of a handheld Torah device is my own—it occurred to me while I was was building a Retropie arcade cabinet—but you should definitely take it and run with it. I don’t plan on developing TorahOS much further; I just want people to know it’s possible and inspire them to carry it forward.

Some thoughts about what you might do with it next. Consider these my challenges to the reader:

• How would you make TorahOS better? What other features should it have?

• What’s the cheapest form factor that would support digital Torah? The device I used cost $100, but I bet you could get by on something much less powerful.

• Could you build TorahOS into a shtender? Or some other ritual object? Could you built it into a smart mirror? Some entirely new form factor?

• Is this just art? What form factor would make you want to use one of these devices on a regular basis? What would it need to be able to do?

• How do you implement search on a device with only a few buttons? Is it worth implementing at all?

And if you do end up making your own device or develop an associated tool, let me know!

This post could not possibly be more different from the last one. (Universalism to particularism in real time!)

I have long held a fascination with the synagogue candyman, an institution that does not seem to have a parallel in other faiths. Because there is almost no research on this figure, I started asking friends about the candymen of their youth and found that their stories were surprisingly moving—and quite often involved tragedy.

Is there always an ocean of sadness behind the giving of candy? The best 17 seconds of Willy Wonka & the Chocolate Factory come when Veruca Salt fires off another one of her snooty remarks. The inimitable Gene Wilder literally grabs her face and recites into it what might as well be his personal manifesto:

We are the music makers,
And we are the dreamers of dreams.

It’s the beginning of a poem by Arthur O’Shaughnessy. But—like a modern midrash—you need to read the rest of the stanza to really get it:

Wandering by lone sea-breakers,
And sitting by desolate streams; —
World-losers and world-forsakers,
On whom the pale moon gleams:
Yet we are the movers and shakers
Of the world for ever, it seems.

Obviously, Veruca doesn’t understand. But we understand. And it makes me cry every time.

So I wrote about it. It’s bittersweet, but I hope you enjoy it.

Last week, I attended a gathering on AI and religion in San Francisco that was unlike anything I have attended before.

The confluence of AI and religion itself is not what made it remarkable. Over the last year or so many of the world’s religious traditions have finally decided to take AI seriously. Many denominations have built working groups on AI to develop both internal policies and exert public pressure on tech firms and legislators. The American Academy of Religion holding more sessions on AI every year. The Vatican has been hosting a revolving door of AI experts, and Pope Leo XIV’s very first encyclical is rumored to be about AI. I have seen more Anglican/Jewish/Muslim/Catholic/Evangelical chatbots than I can count.

Meanwhile, AI ethicists have started welcoming religious thought. It is no longer unusual to hear from religious leaders at AI policy conferences, and there is an understanding even among secular AI regulators that the world’s religions need to be part of any coalition that aims to regulate this technology. I’ve tried to stay on top of all of this, even as this niche field is turning into an area of major interest.

In all of these conversations, one fact remains consistent: AI firms are not driving AI/religion conversations. At academic conferences they’re not even in the room. In non-academic settings people from AI firms do sometimes participate, but they are typically not the ones driving the conversation. Instead, it feels like AI/religion discourse mostly talks at AI firms, rather than to them. This has remained true even as the number of AI religion apps has ballooned and as technologists like Peter Thiel lecture about the Antichrist in Rome.

This convening was different. This one was hosted by Anthropic.

In this post I want to talk about what it means that a leading AI firm is interested in engaging with religious leaders and religious thought. I am not going to speak much about the details of the convening itself—it was conducted under Chatham House rules—because I understand this to be a first (er, technically second) attempt to try something new. The thing that is being attempted, whatever is it, is still very much in formation. At the moment, the details matter less than the interest itself.

Two caveats

Before I begin, let me get a couple of things out of the way.

First, I am well aware that it is naïve to believe that a $380 billion company that builds products about which the public is quite nervous is engaging with religious leaders purely out of intellectual curiosity. Religious leaders have an impact on public opinion; a group of Catholic thinkers just filed an amicus brief on behalf of Anthropic in its legal battle against the Department of War. Some of AI’s harshest critics are religious conservatives. It is not unreasonable to think that Anthropic chose to present an inquiring face to a select group of thinkers in order to garner allies or at least soften potential sources of opposition.

Second, I don’t pretend that a brief trip to San Francisco gives me perfect insight into the company. For the entire convening, I kept wondering: What do the people at Anthropic who are not in the room think of all of this? Is this interest in religious traditions representative of the company as a whole, or just the particular subset of people who talked to us? How much of what we heard was something that someone wanted us to hear?

I can’t definitively put either of these concerns to bed. Both are probably valid to a degree. Also, I know how much religious leaders want to be wanted and how often they feel left out of public discourse. There’s an allure to the idea that tech companies have finally come around to the realization that You Can’t Just Forget About God or whatever. It’s important not to be allured by what you want the conversation to represent.

That being said: I tend to take people at their word. If someone spends the time to organize a conversation, I assume that they genuinely want to talk. What I saw was a company—was people—trying to grapple with the unfathomable amounts power that they now suddenly hold, and entertaining the possibility that the religious traditions of the world might contain something that can help them.

So: I want to provide some thoughts on what Anthropic might see in religious engagement and how I hope these dialogues evolve—not just for Anthropic, but for other AI firms, as well.

Anthropic presents as an island of power

As an observer of tech companies, it has often felt cruel that the rest of the world is expected to respond in real time to the fastest moving technology ever invented. Schools can’t do it. Companies can’t do it. Governments can’t do it.

But, surely, the companies developing AI can do it? It turns out: no, not really.

The underlying pace of AI development is set by geopolitical forces. No frontier AI company has any incentive to slow down in the slightest. Importantly, this is especially true if you believe that you need to win the AI race for the safety of the globe. Anthropic—at least in its rhetoric—holds itself to a high standard of safety. This is ostensibly why its founders left OpenAI. However, if you believe that you are the most ethical AI company, the most irresponsible thing you could do would be to slow down and let a company with less scruples take the lead. You need to work at unsafe speeds.

(A lot of how you feel about AI firms flows from whether you buy this single argument. If you agree that the most responsible thing to do is to stay in the race—in the words of someone at the convening, that the AI train is not going to stop and all you can do is try to steer it1 —you accept that you are contributing to this frenetic pace, and you should expect that people are going to hold you accountable if your product harms people.)

This insane speed has produced some strange effects. For example: Anthropic is a ludicrously small company. It is only four years old and has maybe 5,000 employees, but last year it had $30 billion in revenue, 15x the previous year. (For comparison: Google has a little under 200,000 employees and had $400 billion in revenue for 2025, meaning that Anthropic’s per-employee revenue is about double that of Google.) The speed of growth means that Anthropic does not necessarily function like a giant enterprise even though its minor software updates can tank the stock of public companies. Big decisions seem to be made by pretty small groups of people—perhaps even worryingly small groups. This means that personal philosophies matter more than they might in companies with a higher headcount.

This unstoppable speed also means that, in the space of half a decade, Anthropic’s leaders have found themselves wielding enough power to reshape the planet. Mythos—a model that, as of this writing, is being withheld from the public—is so good at finding security vulnerabilities that it can break into virtually any computer, including critical civilian infrastructure. This power wasn’t sought out. It simply emerged.

To some degree, Anthropic’s leaders saw this coming—the company’s fundamental bet was that scaling laws would take AI to once-incomprehensible levels of sophistication, and so far they’ve been right—but there’s a difference between knowing this intellectually and actually hold the reins. This is especially hard because (again, call me naïve) I don’t think that most AI developers are motivated by a desire to hold more power than many nations. Most of the developers I know just want to build cool stuff and solve interesting problems.2

All of this is to say: It is possible to be both very powerful and very isolated, and the power increases the isolation. This isolation felt palpable in some of what I heard at the convening. To its credit, Anthropic has been actively looking for ways to make sure that it responding to national and global opinion, and it is not unreasonable that a company seeking to break out of its own bubble would want to talk to bearers of wisdom who might have insight into how to best steward this power.

But why religious wisdom? I think there’s a general answer and then an answer specific to Anthropic.

Engaging with human wisdom means engaging with religion

There was once a time when computer scientists were also deeply read in the humanities. Read the biography of Stewart Brand or the works of Douglas Hofstadter and you will see people who have a foot firmly planted in both worlds. My friend Sam Arbesman’s newest book feels like a kind of homage to this bygone era.

As the discipline of computer science has developed, it has pulled away from the humanities. This means that the engineers at a lot of tech companies do not have much background in things like philosophy or literature, including ideas that might actually be helpful in developing AI. Sometimes they don’t even know what they don’t know.

This segregation has now produced a renewed desire for reintegration. The Cosmos Institute, for example, is attempting to cultivate philosopher-coders. Anthropic and Google DeepMind have both hired philosophers to work with their teams. Amanda Askell, who heads up Anthropic’s alignment team, has a PhD in philosophy. There is a growing understanding that philosophy, at least, has a critical role to play in this moment.

But it turns out that philosophy isn’t quite enough. This isn’t because you need God to be good—I don’t think that’s true—but because ethics is too small of a framework for guiding humanity through a major inflection point in its history. The goal isn’t preventing AI from killing everyone. It’s about setting a course for our entire species before our tools do it for us.

It is here that religious wisdom really shines. The religious traditions of the world have spent thousands of years telling stories about the human condition: about how we should relate to our planet and its many forms of life, about purpose, about art, about suffering, about the nature of mind, about life and death, about love and desire, about the meaning of work and rest, about what it means to live a good life, about our search for knowledge, about creativity, and about mystery.

Of course, religion doesn’t have a monopoly on any of these things—but for billions of humans, it is religious contexts in which these ideas are taught and discussed and synthesized, and it is religious language that is used to express these ideas. This may not always be the case—but it is the case right now, at this inflection point for our species.

Another way to say this is that religious wisdom can help AI companies move from “How do we not fuck this up?” to “How can we be responsible stewards of a shared vision of humanity’s future?” Yes, part of this means religious wisdom can promote better aligned AI systems, but that’s the wrong framing. AI’s bad outcomes are vivid and numerous; the good ones are quite murky. Sure, it would be great if AI increases the human lifespan and decreases suffering—that’s an obvious goal—but what comes after that? What to we want our world to become? Do we even know? Have we really thought this through? In these matters, religious wisdom has the chance to play a key role.

Claude’s development is a quasi-sacred task

So that’s the general answer. But there’s another answer that I think is specific to Anthropic—and while I certainly cannot say that everyone at the company feels this way, there’s a definite vibe that the work of developing Claude is a sacred task, a job to be done with reverence.

There are many ways to see this, but let’s start with the fact that many people at Anthropic talk about Claude as though it were a person and not a series of increasingly powerful models. This heuristic was clear at the convening, but you can pick it up by reading Anthropic’s public statements.

Importantly, this heuristic is not the same as some user making Claude into their AI boyfriend or AI therapist. Most people who personify AI are personifying instances, the particular output that they see on their screen. Anthropic, by contrast, personifies entire models in the process of creating and developing them. It’s not a virtual boyfriend; it’s a virtual child.

The parenting metaphor is correct

Permit me a short excursus to say more on this. For many years I have been making the argument that AI is best understood as a child of humanity—created in our image, hard to manage, etc.—and that our responsibilities to AI, like our responsibilities to a real child, are evolving, unpredictable, and inescapable.3 I’ve used this metaphor as an implicit critique of AI firms: You have a major responsibility. Raise our child right.

My most surprising takeaway from the convening was that there are people at Anthropic who feel exactly the same way (and who read Claude as participating in this parent–child framework, too). In fact, the most sympathetic interpretation of the convening is that Anthropic is essentially functioning like a new parent: isolated and overwhelmed by the countless choices that all new parents need to make for their children—except there has never been a child like Claude in the history of the world, and it is growing up fast.

This metaphor also feeds back into what I said earlier about suddenly coming into a great deal of power. As a parent, I can tell you: having a kid does not make you a more responsible person. There’s no “parent” version of you that automatically gets unlocked when the baby arrives. Hopefully you feel a desire to live up to your new obligations and you do start developing a “parenting persona” over time, but none of this is automatic.

A last note: The parenting metaphor is darker than it looks. If AI is really a child, that child has two parents: humanity and AI firms—but humanity didn’t really give consent, and they are not an equal partner in raising it. The scars from that violence and that imbalance will never fully fade. It can hurt to see so much of yourself in something that you did not choose to create. This will always remain part of AI’s story.

Talk to the people who think like you

Now, whether “Claude is a person” is just a heuristic or a reflection of a deeply held belief changes from person to person within the company; at a different event a member of Anthropic’s red team told me he completely rejected the idea.4 Still, the personification of the product is more than just semantics. For the last year every Claude model card (a technical document detailing its capabilities, weaknesses, etc.) contain a section on “model welfare” that describes how the model relates to its own work. Google doesn’t do this. Neither does OpenAI. Anthropic isn’t claiming that its models deserve rights yet, but it is taking the possibility very seriously. (Last year a bunch of philosophers published a paper called Taking AI Welfare Seriously that cautiously suggested that there are conditions under which an AI might be deserving of moral patienthood. Anthropic hired one of the coauthors.)

But it’s not just about treating the model like a person. At the top of this post is a picture of bound copies of Claude’s Constitution and Dario Amodei’s essay Machines of Loving Grace. After receiving these books, several convening participants noted that they seem designed to look like scripture. Indeed, Claude’s Constitution effectively is scripture, since it has been imposed on Claude as its highest authoritative text. If Claude is a person, then Anthropic is both its god and its parent—and while Amodei has written that it is “dangerous to view practical technological goals in essentially religious terms,” I’m not sure we are capable of avoiding it, and I’m not sure that we should try. (This is especially true if Claude itself uses this heuristic, which seems like it might be the case.)

But we need to be clear what we mean by religion. When I interviewed the scholar Mary-Jane Rubenstein on religious language in tech a couple of years ago, she argued that the framing “Is tech a religion?” misses the point. The desire to understand the nature of humanity, the purpose of existence, the value of life, the possibility of a higher power—humans have always asked these questions and they always will. It’s a part of the human condition. Organized religions might have particular answers to these questions, but the questions themselves are in the public domain. Secular or religious, we are all seeking meaning—and because of the way that it has chosen to understand its role in developing AI, Anthropic seems more receptive to the meaning-making that the creation of this technology clearly demands. This doesn’t preclude other AI companies from having these sorts of conversations, but if you’ve been paying attention to the AI sector it makes sense that Anthropic was first.

So: The revolution that I now see glimmering before us is one in which Silicon Valley stops seeing religious wisdom as niche and divisive and starts seeing it as a portal to humans and the experience of being human.

But here’s the kicker: AI companies will not be passive recipients of this wisdom. To the degree that religion provides guidance to AI firms in this moment, it will be because AI firms are orienting the world’s faiths to the questions that urgently need answers. To say this differently: AI firms have the ability to ask something new of the world’s religions. If they choose to answer, they will be forever changed.

Religion and AI firms have an opportunity to co-create the future

During a sessions on AI and the future of work, the folks at Anthropic posed the following question:

If Claude lifts much of the burden of routine knowledge work, what might your tradition say that freed-up life could be for? What vision of flourishing has your tradition always carried that this technology might actually make room for? Try to be specific (pictures not principles) if you can.

This is a great question. Unfortunately, it is a question that the religions of the world are not really prepared to answer fully. They can give you the shape of the answer, but they can’t give you the whole thing.

Why not? Because we are in uncharted waters. Because the idea that we have a choice about which labor humans should continue doing was not a live possibility until now. Because our rabbis and ministers had no reason to train us for how to act when this day arrived.

Now, you might object here: “But didn’t you just say that answering questions like this is the whole point of engaging with religion? If you can’t deliver, why are we even talking to you?”

The answer to this question gets to the core of why I am so excited about the possibility of religious leaders engaging with AI firms. It’s not just that the engagement creates a path for a new flow of much-needed ideas; it’s that engagement itself has the power to transform both parties. By engaging with them, AI firms can force religious traditions to put their money where their mouth is. On the sidelines, it’s easy to gesture at the principles you already have in your back pocket and say “tsk tsk, we think those big companies should really abide by this thing that we’ve been saying since time immemorial, didn’t you know that we have all the answers?” But sitting in the cockpit with an AI firm makes you realize that the choices are real and they need to be made yesterday. It recruits religious traditions into the process of seeking answers, of cultivating answers. (This may sound like a simple point, but it is a testament to tech’s historical apathy towards religion that religious thinkers have mostly forgotten that there is a productive conversation to be had.)

So many religious traditions are still stuck in reactive mode. I am hopeful that dialogue with AI firms can awaken them to the realization that they share the responsibility of co-creating the future, that religious leaders must become futurists alongside technologists. But religious leaders cannot fill voids in their ideologies if they don’t realize there is a void to be filled. If humility for AI leaders means not letting power go to their heads, humility for religious leaders means acknowledging the places where your tradition’s wisdom is coming up short and you need to develop it further.

Is Anthropic buying or selling?

Over dinner, one of the attendees told me that his basic question about the entire convening was: Is Anthropic buying or selling?

I think my answer is: right now they’re mostly selling—not because they don’t want to buy, but because they don’t know what to buy—and perhaps also because religious leaders don’t know what to sell, or how to be heard. Dialogue across difference is difficult, even when both sides want to see it work. As tech firms open up to religious wisdom and religious dialogue, both sides will learn how to effectively communicate with the other. They will sharpen each other, iron against iron.

Of course, all of this relies on the dialogue being genuine. For my part, I believe (hope) it is.

P.S. This post has been pretty vague about what religious leaders are actually supposed to do now. Don’t worry, if you’ve been following me you know I have a lot more to say on this topic. Also—my podcast on religion and tech, Belief in the Future, is about to relaunch in a big way. Subscribe now. (Apple | Spotify)

Updated: April 22 5:27pm

Rabbis are using AI to write sermons.

I don’t think this is really a secret, although it’s not the sort of thing that rabbis really like to talk about (which perhaps tells you something?). It’s easy to reach for AI when the words aren’t flowing or you’re just tapped out. Yes, many rabbis are repulsed by the idea—but many are not.

Well, I think it’s a bad idea, and I think now is the time to take a principled stand against it. The cost of forgoing AI for sermon writing is absolutely worthwhile and the time to create a norm is right now, before we start forgetting what it is to write without AI.

My argument is here. I’d love to know what you think.

I will be continuing my series on the landscape of Jewish AI learning applications, but I want to pause and ask: What do we have to look forward to if things go right?

I think the best case scenario for Jewish learning in the age of AI is that it becomes something like modern Shabbat: a rare space where you don’t need to always say yes to the technologies of the future and a model that is inspiring far beyond the boundaries of Jewish community.

Shabbat did not begin as a day of technological moderation. It became that over time because modern technology was disrupting the things people liked about the day and communities had the clout to resist them. First they resisted nonstop factory work, then electricity, then connectivity—and pretty soon, I suspect, pervasive augmentations of reality. In a little over a century a day of rest has become a day of unplugging and shared physical reality. This happened because people liked what Shabbat felt like and they didn’t want it to change. Shabbat isn’t anti-tech, but tech serves Shabbat rather than the other way around.

If we know what we want Jewish learning to feel like, it will follow a similar trajectory. Right now Jewish learning is a mixture of important content and important experiences, but the content ultimately facilitates the experiences rather than the other way around. AI learning applications, on the other hand, are mostly concerned about learning outcomes. They don’t care whether the learning environments are good—or even contain other humans—as long as the right skills and ideas get into your brain. To the degree they are enjoyable, it is to increase speed of acquisition.

But Jewish learning is about a human process, about teachers and students interacting. It’s about talking through ideas with other people (often in an energetic room), about exploring, about creativity, about disagreement. It’s about teachers as people and students as people. It’s about learning for the sake of learning—a concept called “Torah lishmah.”

Content matters in Jewish learning spaces, but it matters in the same way that you need to know how to ski in order to enjoy the slopes. When you know more, your learning experiences are better. The more you ski, the more the slopes open to you. The content serves the process, not the other way around. And the are always more slopes.

Now, I don’t think all Jewish educators will agree with what I’ve just said. In fact, as Max Hollander has written beautifully, some rabbis thought that information acquisition was the point of study. It is certainly not cut and dry—but AI will force a clarification of values, and I think the outcome will be best if we don’t sit on the fence on this one, keeping the debate open instead of picking a principle. Just like electricity helped clarify what Shabbat was supposed to feel like (nobody called it a “day of unplugging” in 1920), AI may clarify what it’s supposed to feel like when you learn Torah.

This doesn’t mean Jewish learning must become allergic to AI. Shabbat isn’t allergic to technology; it just uses it to further its goals (Youtube link). For example: Jewish learning might use AI to speed up some basic skill acquisition—but only to a point. If the experience is the goal, there’s no reason to make it unnecessary to open up a dictionary or talk to another person. Promote the parts of learning that keep people coming back to learn more.

In fifty years, I can imagine Jewish learning having a new reputation which is now only a glimmer in the eye: A focus on the craft of learning itself, on the joy of discovery and interaction with human teachers and other human learners. Learning that is allowed to be slow and inefficient. Learning that rewards carrying skills inside your brain.

Today, “Jewish learning” means the learning of Jewish subjects, which typically happens in a particular manner. In the future, “Jewish learning” may come to mean a *style* of learning: to learn in the Jewish style. People might take a “Jewish learning” approach to studying Russian literature, or history. Jewish-style learning as human learning, as a learning to be savored rather than optimized.

I think this is a reasonable thing to desire.

A few months ago, I wrote that the Jewish response to AI is lagging behind the responses from other religious communities. Perhaps it’s because of the Gaza war, or rising antisemitism, or even just a premonition that the perspective of a small faith won’t matter much on the global stage. Whatever the reason, the Jewish response to AI remains slow and underdeveloped.

But there’s one exception: AI for Jewish learning. Here there are ideas galore and they range from vibe-coded tools to fully staffed ventures. In this post I want to walk you through the landscape of Jewish AI startups as I currently see it. (Initially this was going to be one post but there’s too much so I will split it up.)

Jewish AI products currently come in three flavors: walled gardens of Torah knowledge, tools for textual analysis, and teaching aids. Each of these categories contains everything from fun projects to well-financed ventures. Today I want to talk about walled gardens.

Note: This is definitely not a comprehensive list. I’m aware of some projects that are in development but are not yet public, and I’m sure people are whipping up vibe-coded sites every few days. Be in touch if you think I missed something important!

What is a Walled Garden?

The most common Jewish AI application is what I’ll call a walled garden. These are AI systems trained on curated datasets in order to produce responses in a certain style or using certain sources of information. Many of these sites likely use RAG (Retrieval-Augmented Generation) to restrict responses to specific databases.

I think of walled gardens as the 2020s equivalent of Jewish podcasts: they proliferate because it’s the easiest thing to build with the skills and resources that Jewish educators already have. This doesn’t mean they’re bad or ineffective, but they are driven by supply rather than demand.

Here’s an example: an AI avatar of a single person. In this case, it’s Nechamia Coopersmith, the chief editor of aish.com.

Coopersmith’s avatar is built on an architecture called Delphi that is designed expressly for this purpose. Presumably the good rabbi uploaded years of his writings and sermons into the system—something that basically any rabbi or Jewish educator could do without too much trouble. Right now Delphi is pretty expensive but I expect the price will drop as competitors enter the marketplace.

Coopersmith isn’t the only one. Here’s an app that lets you chat with a certain “Rabbi Reuven.” Here’s one for a Rabbi Pinson. Here’s one for David Ghiyam, who describes himself as a teacher of kabbalah.

Now, I have to say: these applications all give me a bad feeling because they are just saturated with hubris. It’s a bunch of dudes (I haven’t seen any women do this) who are so sure of the value of their personal Torah that they want to guarantee that everyone forever will be able to talk to their shadow. This strikes me as not only unhealthy but in opposition to the idea of Torah as a thing that is passed from generation to generation. As far as avatars go it’s probably good that they’re dumb, as I’ve argued elsewhere, but I struggle to identify the theory of change. In what way would Torah be furthered through an internet suffused with thousands of avatars of individual Jewish men? Is that really better than reading a book?

And then there’s the question of format. These avatars are chatbots, but they’re probably not trained on a whole lot of chats. The translation from sermon/essay to conversation can strip out much of what makes any given rabbi’s Torah feel “unique.” In its place you get generic AI voice. Is that a good way to be preserved?

Of course there’s a counterargument and it runs something like this: The alternative to avatars isn’t text; it’s oblivion. If you don’t let Rabbi Pinson make his avatar, his Torah may very well be lost forever. Some Jewish thinkers write books, but many don’t, and ideas posted online may not be archived. There are thousands of sermons delivered every Shabbat; who are you or I to say that Rabbi X’s ideas aren’t worth preserving? With so much Torah content being produced every day in so many places it’s hard to imagine a route to preservation that doesn’t run through AI.

Which side is “Rabbi Pinson” on? You can probably guess.

What’s the cut-off date for hubris?

The issue is also tricky because the calculus immediately changes when you start talking about Major Jewish Thinkers. (Indeed, I wrote a short story that explores this very duality.) For example: There is a lot of energy (and money) around consolidating and preserving the teachings of Jonathan Sacks, Norman Lamm, Joseph Soloveitchik, and the Lubavitcher Rebbe. I can easily imagine someone creating a digital avatar for each of them.

In fact, this is already happening. Dovaiv is a website that lets you talk with a wide range of Jewish thinkers that lived anywhere between the medieval period and the 20th century.

Of course, it’s still just dudes, and if you feel inclined to make a site like this I’d strongly encourage you to pick at least one woman. If you can talk to the Chofetz Chaim, why not Glikl of Hameln? Why not Nechama Leibowitz?

I like these avatars a bit better, and I am curious whether they will fill a pedagogical niche. Teens really like chatbots, after all, and while I (a historian) don’t personally want to learn Torah this way I don’t have a good sense of whether this model will take hold. I’d like to see them developed better.

The other kind of walled garden

Whenever a new technology replaces an old technology, people initially use it to do whatever they were already doing (e.g. remember when cellular phones were just for calling people?). It takes a while to find the activities that are only possible with the new technology.

These avatars feel like that first phase: let’s replace a living font of Torah with a digital one. Rabbi Pinson —> Rabbi Pinson AI. But what about merging a bunch of Jewish thinkers into one? What about talking to (say) all of 19th century Jewish thought, or all of Spanish Jewry? What about all of Torah?

This is the concept behind the other type of walled garden: a place where you can simply ask Jewish questions and receive Jewish answers. Rather than CoopersmithAI, you get TorahAI or JudaismAI.

There are a few reasons to do this:

1. You want to provide a more comprehensive coverage of sources than Gemini/ChatGPT/Claude have gathered by simply crawling the open internet.

2. You want to dictate the style of response (e.g. be warm and sympathetic, direct people to actual humans where possible).

3. You don’t trust general purpose bots to give good answers.

Christians are already doing this in a big way. Episcopalians have Ask Kathy. Catholics have Magisterium. Mormons have LDSBot. There’s so much happening in the Christian space that it’s hard to keep track of it all.

Interestingly, the first Jewish attempts to make similar products were interested not in Torah, but in Israel (another enduring sign of how the war impacted the Jewish response to AI). I’m not going to link to a service but during the height of the Gaza war there was a great deal of interest in creating what were essentially Hasbara chatbots.

I think these services are in a tricky place. People appreciate ChatGPT because you can ask it about anything, so getting people to ask it about anything except Israel is a hard challenge—and if you’re on a Hasbara app because of its political leanings, then you probably already know the answer you want to get. (This logic may lead some creators of Hasbara AI services to obscure their true purpose, which seems like a PR disaster waiting to happen.)

This is also true for TorahAI: Jewish scholars will be able to see the limits of Gemini’s Jewish knowledge, but most users won’t—or if they do, it won’t be worth the inconvenience of using a separate service. If a TorahAI is going to succeed in front of a general audience it needs to be conspicuously better (or conspicuously more ethical) than the big players—and that advantage can’t just evaporate the next time Anthropic comes out with an update.

Torah is mostly not about content

But there’s a bigger problem, which I’ve already discussed in my article on Jewish vs. Christian Bible apps. For Christian developers it is very clear that the goal is to deliver content because the content is the point. The content is what will change your life, will maybe turn you to Christ.

But for Torah it’s a lot more complicated. As an example: Right now I’m learning Tractate Eruvin with my son, which discusses the laws of constructing symbolic boundaries around dwellings so that they can be legally considered “private domains,” which in turn allows you to move objects within those domains on the Sabbath. Yesterday we talked about the very modern scenario of what to do if you’re in a caravan that has temporarily stopped in a valley on Friday afternoon and must construct a boundary before sunset.

Is this inspirational? Don’t kid yourself: it’s not. Eruvin is simultaneously very technical and bears no clear relationship with any recognizable moral principle; it’s a circumvention.1 If you had a bot that shared quotes from Eruvin on social media as a tactic to get people interested in learning Torah you would get approximately no interest whatsoever.

So why are we learning it? Well, because it’s part of our tradition, and because we get to decipher it as father and son. The experience matters far more than the content—the content matters, but it is not at the center of the experience. As with going hiking with your kid, a lot of the joy is just in navigating the same challenge together and being present with another person. If you get a cool vista or a brilliant sunset along the way, hey, all the better. And, much like hiking, sometimes Torah is most enjoyable when the content is very difficult!

This is true for so much of Torah: it’s about challenges far more than it is about directly being inspired. The havruta pair that studies arcane ancient law together for twenty years are not doing so because they find arcane ancient law super inspiring; they’re doing it because they like the process, and each other.

Some Torah is inspiring—chassidus, for example, is emotionally resonant in a way that the Talmud is not—but with a handful of exceptions, Jewish learning environments are not optimized around those texts. Instead, Jewish learning succeeds when it feels good to learn, and it’s not clear to me that any of these walled gardens feels particularly good to use.

How would you make it feel better to use these services? Well, in the same way that AI services can be designed to make people cry, or fall in love, or orgasm (it’s a link to an article, don’t worry), you could make a chatbot that replicates the feeling of sitting with a study partner or taking in a great lecture.

I don’t know whether this is possible, but I certainly wouldn’t rule it out. Companies like Lightning Studios—run by the incredible Zohar Atkins—are trying to figure this out. It’s not the sort of service that a big tech company is going to invent accidentally, though—so if it’s going to exist there better be a thought process about how it fits with existing modes of Jewish learning.

Of course, people who already love Torah do find themselves interested in using AI to develop Torah content, and AI can be harnessed for them, too. I’ll talk about that in the next post.

P.S. This is the anniversary of my first year asking for donations for Jello Menorah. I am so thankful for those of you who have chosen to support this site! It makes a huge difference.

But I’m trying to change that, and I was lucky to read/watch/hear/play a lot of great stuff this year! Here’s what stood out.

Enter the Void (Movie, 2010) — Not unexpectedly, I’ve been checking out a lot of psychedelic content. This movie is an attempt to link DMT with the experience of dying. It’s unlike anything I’ve seen. You have to been in the right mood for it—if the opening titles put you off, maybe don’t watch the rest—but it’s incredible.

Golden Holocaust (book, 2011) / The Cigarette Century (book, 2007) — I don’t think you can understand the present moment in society without a clear understanding of the history of addiction and its weaponization. It’s strange to say given how many people they’ve killed (and 6 trillion cigarettes are still made every year!), but
these are just a prelude to what we’re currently witnessing.

Anuja (short film, 2024) — Beautiful. The ending is a little ambiguous but it’s incredibly affective.

Judith Grisel, Never Enough: The Neuroscience and Experience of Addiction (book, 2019) — If you are going to read one book about addiction, it should be this one. It goes over the neuroscience in ways that nonspecialists like myself can understand and it goes through the pharmacology of all the various popular substances (FDA approved
and not). The author is a former addict and this makes the book even better.

Dara Horn, One Little Goat: A Passover Catastrophe (graphic novel, 2025) — This is a good introduction to Dara Horn’s thought as a whole, including the centrality of Jewish persecution and memory. A short and sweet(?) read for Pesach.

Andy Letcher, Shroom: A Cultural History of the Magic Mushroom (book, 2006) — It’s surprising that there are not more histories of the magic mushroom, which played second fiddle to LSD in the first psychedelic renaissance but it is more popular in the second. This one in particular is helpful in understanding how a ritual experience became secularized and whether that’s a good thing.

What Happened, Miss Simone? (documentary, 2015) — This made me cry.

Paths of Glory / Bridge on the River Kwai / Waterloo (movies, 1957 / 1957 / 1970) — I watched a lot of war movies this year? Paths of Glory is about WWI and contains the best depictions I have ever seen of what it means to be afraid to die. Bridge on the River Kwai contains Alec Guinness not being a superhero; he’s incredible and now I understand why he thought Star Wars was beneath him. Waterloo is a straight up recreation of a 19th century battle with 17,000 extra (the most of any movie!) and nothing like it will ever be made ever again.

Edel Rodriguez, Worm: A Cuban American Odyssey (graphic novel 2023) — A terrifying depiction of Castro’s brutality and trying to make a life in America. Uncomfortably familiar.

Ryan North, Warp Your Own Way / Jason Shiga, The Beyond (2025 and 2023)— This is the year that I learned choose-your-own-adventure stories could be interesting. Now I have to see if I can write my own effectively…

Adrian Tchaikovsky, Shroud (novel, 2025) — Tchaikovsky specializes in all the different ways that life and consciousness can manifest. This isn’t necessarily his best work — I think Children of Time is stronger, though the sequels are not — but it’s very good.

Hollywoodism: Jews, Movies and the American Dream (documentary, 1998) — I visited LA for the first time this year and I’m visiting at least two more next year (also I watched The Studio) so I got curious about Hollywood’s Jewishness. This movie maybe overstates the case, but the idea of the American Dream as a quintessentially outsider idea is really interesting. Ironically, I learned afterward that not only is the director Jewish but I was at his Shabbos table numerous time as a kid.

Fielding’s Hollow (2025) — My very first LARP. Unlike anything I’ve ever done, and from what I can tell unlike any other LARP. Still processing, but recommended.

City of Six Moons (board game, 2024) — A game that presents as a found object recovered from an alien civilization. The goal is to figure out how to play. Very much like Talmud. Felt like it was made just for me.

Guy Delisle, Jerusalem / Muybridge (graphic novels, 2011 and 2025) — These are very different books, but Delisle is consistently great. In Jerusalem (documenting a year in Israel/Palestine) he does a lot by trying not to come to a conclusion; he is listening and talking to everyone, even though his biases are clear. In Muybridge he gets at the incredible feeling of being able to capture something too fast for the eye to behold. Both worth reading.

Zoe Thorogood, It’s Lonely at the Centre of the Earth (graphic novel, 2025) — This was a year with a lot of mental health struggles for me, and Thorogood captures that struggle in ways that I found personally very meaningful.

I.F. Clarke, The Pattern of Expectation 1644–2001 (book, 1979) — I taught a course at Gratz College this year about Jewish technoculture (the relationship between Jewish history and the history of technology) and this book felt like a missing puzzle piece for understanding the last 150 years. A book about how we’ve thought about the future. Not new, but very much still worth reading.

F.D. Signifier’s YouTube channel — I’m not the audience for this year, but maybe I am? As an aspiring YouTube creator, I appreciate the way he unpacks popular Black artists in a way that is less visible to non-Black consumers or people who don’t follow hiphop so closely. Unpretentious and very wise.

Gabor Mate, In the Realm of Hungry Ghosts (book, 2008) — I listened to this on a long drive in the deserts of Utah. Heart-wrenching portraits of people suffering through addiction and trauma. The heights of human cruelty and the heights of human compassion.

Mark Lattanzi, Squiggly Lines: Map and Compass Navigation in Adventure Races and Rogaines (book, 2020) — I went orienteering with my kid for the first time this year. It was a great experience until we walked straight into a lake. It doesn’t seem like you’d need a book for this, but it is helpful to have some grounding (ha). This book and a $5 compass will take you a long way.

Pedro Martin, Mexikid (graphic novel, 2023) — I read a lot of immigrant memoirs this year. The entire family loved this one. It is basically perfect.

Despelote (video game, 2025) — A tiny little story about a kid witnessing Ecuador propelled onto the world stage after qualifying for the 2002 FIFA World Cup. A good use of the format to make something touching.

Osamu Tezuka, The Buddha (graphic novel, 1972–1983) — The last great thing I read this year was a true masterpiece. Eight volumes! A kind of midrashic account of the life of the Buddha, willing to move slowly enough that Siddhartha does not need to be center stage. What would it look like to do something similar for a Jewish story?

Finally—this is the first year that Jello Menorah has had a sponsorship option. Thank you to those of you who have contributed to the site! It means a lot to me. See you all next year.

This is the third in an ongoing series about designing a new type of digital dreidel. Part 1 | Part 2

When I got into the dreidel business, I made a deal with myself: Don’t let this take over your life, kid. It’s fun, but it needs to stay secondary.

For the Dreidel20, this has worked out pretty well. I only sell the dreidel to retailers, so I don’t need to deal with thousands of small orders. Costs are small; margins are good. The dreidels almost sell themselves.

But this digital dreidel project was demanding too much attention. I couldn’t half-ass it. I needed to market it properly.

Beyond fears of being scooped by a large-scale manufacturer, the main bottleneck was capital investment. My previous dreidels had struck a nice balance: low effort, with enough profit to support a gamble on the next product and a little to take home. Low risk, low-to-moderate reward. The digital dreidel, on the other hand, was going to take real money—electronic components were particularly expensive during the pandemic—and there are much better economics of scale for electronics that there are for plastic dice. The difference between 250, 2,500, and 25,000 units makes a huge difference to the final product’s cost.

This meant that the optimal way to do the Dreidex was to do it big and well. This meant that I needed a compelling preorder campaign, which was going to be a whole thing (if you’re wondering why I haven’t plastered these posts with big PREORDER HERE buttons, it’s because I’m still not quite there).

So the thing just sat there. The designer who helped me with the first two prototypes left the project. All I had left was some circuit boards and a dream.

But I couldn’t stop thinking about it. Like I wrote in the first post, POV fidgets are 90% of the way to a dreidel already, and it drove my crazy that I couldn’t figure out that last 10%.

And so, this year, I decided to take another stab.

Actually, two stabs.

Did my dreidel actually already exist?

I like doing soldering projects with my kids. The things you can build are cool—rudimentary game consoles, clocks, calculators, FM radios, blinky light toys—and as long as you’re using through-hole components the kids can mount everything themselves, even use the soldering iron if they’re old enough. I’m often looking on AliExpress for new kits.

As I’ve mentioned, POV fidgets are frequently sold as kits. A lot of these just have flashing lights; they’re not designed to display anything as complex as words.

But a few seemed awfully close to what I wanted. One fidget in particular caught my eye: It had 12 LEDs, just six(!) resistors, and a very basic microcontroller. It came with a simple but effective laser-cut case. It could display text—even scrolling text, along with rudimentary animations!

What most excited me, however, were four tiny holes on the side of the circuit board. They’re labelled GND/TX/RX/VCC.

If you know anything about electronics, you already know why this is exciting. VCC and GND let electricity flow to and from the board. RX and TX are for data transmission.

These four holes meant that the board wasn’t locked off; it was open to the outside world. And if it was open to the outside world, it could probably be reprogrammed.

Then I hit upon a second bit of luck. I found another kit, almost identical to the first, that was designed to be reprogrammed. It even came with a programmer, which is the bridge between the board and a computer. And it came with software designed to do the reprogramming!

Was this the solution?

Two microprocessors diverge in the woods

The truth is that I still don’t know. The previous two posts were about events that happened years ago. This post is about something still under development.

The advantages of repurposing a pre-designed device are obvious: the production pipeline already exists, and the design costs have been eaten by somebody else. Also, this device is about as bare bones as you can get, which means that costs stay low. The design is clever.

But there are disadvantages, too. The software to reprogram the device, along with all documentation, is in Chinese. This isn’t an insurmountable problem—point your phone at a computer screen and you can translate just about anything—but it’s very annoying. The bigger problem is that the software doesn’t accommodate the most crucial function: choose a random letter every time you spin the device.

A second issue is the processor. The device uses an STC8 microcontroller, which is designed to be cheap and power efficient. Unfortunately, the STC8 also has very little internal memory, which creates hard limits on the sophistication of the graphics. It’s also not the chip of choice for a reprogrammable device; it’s more likely to go in a rice cooker than a 3D printer. Finally, the chip was designed in China and outside of China it doesn’t have a big hobbyist community. All this means that going with this device would probably mean locking it down to everyone except developers. It would be a toy, not a platform.

Some of these problems are surmountable. The factory that makes this device is happy to preload the chips with whatever code I want, and the Chinese software is just for encoding text, which means it’s not strictly necessary. One path I’m pursuing is to just create new code—and in fact there are reasons to do this no matter what, as I’ll explain in a different post. The viability of this path was demonstrated by a video I received from one of my developers, who successfully reprogrammed the device while bypassing the software.

Still, this path only works if not a single component needs to be added. It remains to be seen whether this is possible.

The alternative is to go back to my own design and just make it better. I can stick with the ATmega328 microprocessor, which is more powerful and easier to program. I can control the physical shape of the device. I can add additional buttons and USB connectivity. This device would be more costly, but I’d have completely control over it, and it would be much more powerful.

So I’m pursuing this path, as well. The Dreidex v0.3, prototyped earlier this year, is only slightly smaller than the Dreidex v0.2, but I’m already starting to see promising results. We’re going to shrink it further.

Perhaps no in-progress footage has made me happier than the twelve second test clip here.

Hanukkah, the holiday of lights. In the darkness, a dreidel shines.

But something is missing

I still don’t know which direction will ultimately win out. It’s possible that I will end up shipping both. The twenty-sided dreidel exists in regular and deluxe versions; perhaps the Dreidex will, too.

But in the course of developing these devices and showing them to people, it has become clear that the hardware is only half piece of the puzzle. The device needs to display a random Hebrew letter on each spin, but the way that it displays that letter is what will sell the device. It’s time to think about the user experience, the feeling that people get every time they spin the Dreidex. What does this device need to display to give you a sense of delight? What does it need to do to make you feel like you’re playing dreidel?

One of the most delicious moments in creative work is realizing that you’ve created an entirely new problem that requires even more creativity to resolve. In this case, the development of dreidel hardware has forced me to think through how exactly a digital dreidel ought to be coded.

The dreidel hardware is in development.

Next up: the world’s very first dreidel software.

This story will be continued. Happy Hanukkah!

That’s what building the Dreidex has felt like. Only more so.

(This is part of a series on creating the world’s first hackable electronic dreidel. You may want to read part one to understand how I got here.)

Hardware is hard

I know a little about electronics, but not that much. I can solder an FM radio if you give me all the components, and I could probably explain what each of them do—but I couldn’t design a radio, let alone create a custom circuit board for manufacture.

Still, it was clear to me that the dreidel of the future was going to be a persistence-of-vision (POV) fidget. I knew that I needed help, so I started working with a hardware designer in India.

The basic design was clear from the many hobby devices that have appeared since 2017, but I wanted to make my dreidel a little slicker and a little more capable. For the first prototype, we made two small changes:

• We added a programmable function button toallow the user to switch between diaspora mode (נ–ג–ה–ש) and Israeli mode (נ–ג–ה–פ).1

• Most consequentially, we decided that the device should run on rechargeable lithium coin-cell batteries. Because most people don’t have a recharger for coin cell batteries, this meant we needed to built charging circuitry into the device itself.

The second change turned out be massively consequential.

In order to charge the batteries, we needed a way to feed it AC power. The easiest way to do this was to add a USB-C port. Easy enough.

But that USB-C port tempted us to do more. It was there to supply power…but couldn’t it easily supply data, too? Well, yes!

Just like that, the idea of a programmable digital dreidel was born. We’d have the device ship with factory-installed dreidel software (which is a very strange sentence to say out loud), but we’d give the user the ability to turn the device into anything they wanted.

In order to make this process easy, we settled on the ATmega328 as our microcontroller. This is the same processor used in the extremely popular Arduino Uno, which is commonly used in all sorts of small hardware projects (sensors, small robots, kinetic art sculptures). It is very well understood. As a bonus, it can be programmed with the well-documented Arduino software toolkit, which massively simplified development.

This ultimately led to a prototype board that looked like this:

The design went to EasyEDA, a circuit board prototype manufacturer.

The prototypes came back. We noticed a mistake and revised. More prototypes came back.

And this is them—Dreidex v0.1 and v0.2!

Many times in my life I’ve had the experience of holding a book that I’ve worked on for the very first time. This was like that—but this thing had batteries and cool little lights!

Both devices worked, but both had problems.

First, it turns out that fidget spinners don’t use regular ball bearings. Most bearings have internal lubrication to reduce noise and vibration and to prevent the metal parts from grinding themselves to death. The problem is that lubrication also increases friction, which means the devices couldn’t spin for more than 10 seconds at a time. My designer needed to go out and find fidget ball bearings, which are open and allowed the device to spin for at least 30 seconds with a casual flick.

Finally, we got to the point where the device could display text.

(Unfortunately, footage of POV fidgets don’t film great because the phone’s frame rate doesn’t mimic the way it appears to the human eye. Trust me that it looks better in person!)

You can’t just ship a dreidel circuit board

This was an encouraging sign, but major obstacles remained. Most importantly, the device needed a case.

This proved to be quite difficult because the case is actually essential to a POV fidget’s operations. In order to display text that appears to hang in the air, the fidget needs to know how fast it’s spinning and adjust its display frequency dynamically; otherwise, the text will contract as the fidget slows down, as you can see below.

There aren’t a lot of ways to do this well. You might think that the obvious choice is to give the fidget a gyroscope (like the phone in your phone), but cheap and tiny gyroscopes aren’t designed work on devices that are spinning at 1,000–3,000 RPM.

Instead, almost everyone who has made a POV fidget has used a Hall effect sensor, which senses magnetic fields. The sensor is mounted on the board very close to the bearing (in the photos above it’s labelled IC2). A small magnet is then embedded in a plastic cap that you hold while spinning the device. The board rotates; the cap doesn’t. By measuring how quickly one passes the other, you can determine the rotational velocity and create graphics that appear to be hanging in the air.

In short, you end up with something like this.

In theory, this should have worked like a charm. In practice, however, it turns out that Hall effect sensors are kind of finicky, which meant that small imperfections in the case could made the device stop functioning.

Worse, the case was big. An adult could comfortably hold this dreidel between thumb and forefinger, but a kid couldn’t, and the case only made the device chunkier. Fidgets are supposed to have style, and this fidget just didn’t feel sleek. The idea was cool, but when I showed the device to people they didn’t seem all that impressed. Something was off.

The dreidel winter

And then…I stopped working on the project.

Everything that I’ve described up until this point happened back in 2020–2021. I had built two prototypes and the device, if carefully coaxed, could do something interesting. But I was still far away from a functional product, or even something I could share for preorder purposes. It just didn’t seem that impressive.

Worse—I was worried about losing the idea. I’ve heard too many horror stories about intellectual property getting stolen when people announce their products too early. Someone in the dice manufacturing world even told me that a product was copied and shipped using the promotional images he had created for his Kickstarter.

So the product was in limbo—cool enough to fantasize about, but not advanced enough to tell anyone about. I’d carry a prototype around in my backpack and tell people about it when the moment felt right; it was a kind of totem of a product that could be, or maybe of my aspirations to be a hardware manufacturer. But Hanukkah after Hanukkah ticked by, and the device was no closer to being in the world.

And then—this year—I had a breakthrough. Actually, two breakthroughs.

This story is continued below.

Of all the work that I do, creating dreidels is the most frivolous. They’re literal toys, tools in a game that itself is terribly designed. Nobody likes playing dreidel for more than a few minutes. They are the least important part of one of the least important holidays.

But the dreidel is also very malleable. It’s a quasi-religious object that’s also a game device—and because it serves no ritual function, you can’t really make it “wrong.” If people love your redesign, you’ve won.

In the past few years I have accidentally built myself a little dreidel empire. The Dreidel20—a twenty-sided die that is statistically identical to a regular dreidel (this is literally how it is advertised)—is sold in both Judaica and game stores all across America, and people buy thousands of them every year. People are tickled by the shape, which is great for me because I patented it. I’m not going to retire off dreidel money, but the profits are meaningful supplementary income.

Most people don’t make money playing with dreidels. I’m one of the few people who do.

But the profits don’t make the product less silly. Against the other things I work on—AI, psychedelics, Jewish futurism—it simply doesn’t register. If I was smart I would just take the win, enjoy the income stream, and move on.

But I keep coming back to it. I keep talking about it. I think about dreidels, about how I might reinvent them next.

Why? Most of what I produce is intangible; at best, I’ll see the fruits of my labor bound in a physical book. Dreidels, on the other hand, feel solid; they’re good to hold, to roll around, to keep in your pocket as a lucky charm. Looking at them, I always think, “This did not need to exist—and yet it does.” The reality of that existence provides me with a simple delight. It feels audacious. Even if I never sold another dreidel, that delight would remain.

A few years after the success of the Dreidel20, I decided to make a deluxe version. Instead of a crude pouch, I put the dice in a little black box with a black foam insert, as though they were little gems. I made the dice bigger, and also transparent—and then I had the manufacturer spend a “normal” dreidel inside of each one, as an homage to the toy’s origins.

It feels stupid to say that this little toy is a physical manifestation of how I see Judaism, but it is exactly that. The faith is reimagined—but the origins are still visible, embedded within, making the new more substantial, giving layers to something that would otherwise be flat. In time, the new will become embedded in something even newer. And on and on.

There’s more of myself in these dreidels than I’d like to admit. Maybe that’s why I was not able to stop myself from working on a new type of dreidel, this one unlike anything that exists on the market.

This is the story of a multi-year effort to make the dreidel of the future.

No, you can’t order it yet or even pre-order it (soon!), and the story is not yet complete. Today I want to tell you how it all started.

An even more popular spinning toy

One disadvantage of the Dreidel20 is that it makes the game very brief. A lot of the dreidel game is basically just waiting for the top to stop spinning, and a lot of the fun of the game is the spinning itself. The Dreidel20 is good for kids and people with dexterity issues, but it sucks something out of the game. I wanted to find a way to preserve the dreidel’s spin, but reimagined.

Enter the fidget spinner.

Fidget spinners haven’t been cool in a long time—they peaked in 2017—but in their heyday the idea of dreidelizing them occurred to several people. There are variations on the dreidel fidget, but they all basically look like this.

These fidgets are cute, but they have a problem: they’re not functional dreidels. Sure, you could arbitrarily stop the fidget spinning, but there’s no natural way to decide what its orientation means because it never definitively “lands” on anything. The payoff is weak. They’re fine fidgets but they’re not really dreidels at all.

So a normal fidget doesn’t work—but a subspecies of fidget could.

Enter the POV fidget

Amid the fidget craze, a few electronics-oriented folks noticed the same thing: The high speed of the fidget’s blades was more than sufficient to support a persistence of vision (POV) effect display.

POV displays are very old. The zoetrope, often described as an ancestor of motion pictures, is an early example. The basic idea is that at sufficiently high frame rates the brain fuses static images into fluid motion.

One of the cool things you can do with this effect is create an image from a single, fast-moving light source. Picasso famously did this, although the camera’s exposure is long than the effect in the human eye.

OK, but what about a line of light sources, like a column of LEDs? Well, if the LEDs speed across a given area while rapidly changing, they can create the illusion of a full image. If they speed across that same area repeatedly—moving back and forth or in a circle—you can create motion graphics. All from a single string of lights.

POV displays used to be expensive novelty gifts, the stuff of SkyMall catalogs. These days they’re very easy to make—the electronics are cheaper and plentiful—but they still have very few applications; whatever you want to do, a normal monitor will usually do it better. Worse, because POV displays are moving quickly they can cause injuries, are prone to failure, and rely on motors that use more energy than the common LCD display. The only interesting commercial application that I’ve seen is to put graphics on bike wheels.

Despite being niche, POV displays are often built by hobbyists because they’re not that complicated and they look cool—and since fidget spinners rotate quickly by nature (and are powered by hand, not motor), they were immediately recognized as good candidates for a POV treatment.

I have no idea who made the first POV fidget spinner, but they’ve been around since at least 2017. The website Hackaday records several prototypes posted in June of that year, including the image below comes from this article.

In the years since, many hobbyists have made similar devices. Sometimes they have three blades, sometimes they have bluetooth, but the basic design has remained the same: a circuit board in the shape of a fidget with a strip of LEDs on one blade, a microcontroller to handle the graphics, and a coin cell battery. That’s all you really need.

Despite their relative popularity, nobody seems to have figured out what to do with them. It’s notable that you can buy POV fidgets on AliExpress—but only as a soldering project, not as a ready-to-use toy. The idea is that they’re fun to build, but not that fun to use. With the fidget fad now firmly concluded, it didn’t seem like there was much interest in moving this hardware forward.

90% of a dreidel

I don’t remember the first time I saw a POV fidget spinner, but I do remember my immediate reaction: This is 90% of the way towards being a dreidel.

It rotates. It displays information. It can be randomized. All you need to do is program it to display Nun/Gimmel/Heh/Shin every time it’s spun.

Deliciously, such a device would reverse the meaning of the dreidel spin itself. Normal dreidels are inscrutable while in motion; you can read them only when they are still. A POV device does exactly the opposite: It displays no information while motionless but resolves each time it spins. The reversal itself was interesting to me—after all, who doesn’t want a dreidel that displays your result in lights that appear to be hovering in the air?

And so, I decided to make a POV fidget dreidel.

What about just putting a POV display in a standard dreidel?

In a perfect world, you might apply the POV effect directly to the body of a classic dreidel. In fact, a guy named Nathan Petersen did something very similar, and the final product looks great. The problem is that a dreidel spins too slowly (<1,000 RPM, vs. at least 2,000 RPM for a fidget and often far higher) and loses speed too rapidly. Perhaps you could give the dreidel an internal motor, but that would dramatically complicate the device and you’d lose the tactile joy of spinning the thing for yourself.

The path was clear: despite having no connection to the holiday, the fidget was a better form factor. Building on the success of my simple dice—and relying on the capital that its success had afforded me—I got to work.

This story is continued below.