David Svi Kalman (7:30)

It's great to be back.

Yehuda Kurtzer (7:32)

Yeah. Let's start at the beginning. What are we getting right and wrong in the conversation about AI? Let's start in the Jewish community. And then we can kind of branch out and figure out whether the Jewish conversation is at all different than the kind of global public conversation about the ethics and limitations of AI.

David Svi Kalman (7:52)

Yeah, so I would say the Jewish conversation right now, to me, feels a little bit behind. AI has been a subject of religious interest for around three years since the launch of ChatGPT. And I am seeing now lots of religious communities, Mormons, Muslims, Protestants, Catholics, who are taking this stuff seriously at a kind of organizational level. I think the Jewish community is going to get there, but it's not quite there yet. And the kind of obvious reason is Jewish thinkers in the last few years have really been occupied with something else, with the war and its fallout and anti Semitism and all those things. So this has kind of taken a backseat to a lot of those. That doesn't mean nothing has happened. And I think a major focus has and continues to be on the use of AI within Jewish educational contexts. In the same way that when social media launched, there was a lot of interest in using social media to distribute Jewish ideas. And when the Internet came out, there was this idea of like, hey, maybe we should send around divore Torah, we should send out sermons by email. So it's another one in a long line of broadcasting mechanisms, tools to kind of get the word out or to make Jewish ideas more accessible to the public. And that's all great. I think that's really important. What I worry gets missed is when you use a tool for the reasons that its creators want you to use it, you know, use it to broadcast, use it to develop content, it makes it harder to see the ways in which it is problematic or to be critical of it. So in the same way that Jewish leaders, I think, failed to be sufficiently critical about social media when it came out, I think there is a real concern that Jewish leaders will fail to be sufficiently critical about AI because it's like, well, it's so useful for my, you know, to develop my finances, to write my sermons or things like that. So there is a kind of small use case, and I'm hoping that there's a larger perspective that develops and emerges.

Yehuda Kurtzer (9:39)

I have a lot of things based on what you said that I want to dig into. But what would you characterize as those risks that you think that the Jewish community is not seeing because it's primarily focused on utilizing the tool more effectively for the things that it wants to get out of it?

David Svi Kalman (9:52)

Yeah. So I think the risk the Jewish community faces are similar in a lot of ways to the risks that everyone else faces. Concerns around dehumanisation, I think for me are really primary. The fact that AI can replicate people's behavior but at the same time does not get treated like a person. That's kind of like the entire value proposition of AI means that it's very easy to devalue educational labor, all kinds of text production, research, whatever it is. And there is going to be an attempt in the next few years to basically look at every single job that human beings do and say, can AI do this? And we'll try it and sometimes it'll work and sometimes it won't, but the times that it works, maybe it shouldn't work, maybe we shouldn't have gone down that road. So there is that kind of larger concern. There's also more specific concern around antisemitism. I would say this is something we saw a few months ago when X's Grok AI started talking about Mecca, Hitler and praising the Holocaust and doing Holocaust denial, all kinds of things like that. You know, some of that has to do with the particularities of X and Elon Musk, but some of it has to do with AI's ability to be manipulated or to be biased towards different media in various ways. And because the Jewish population in America and around the world is pretty small, most people learn about Jews through media. And if that media becomes saturated by AI, then AI's opinions about Jews ends up playing a kind of outsized role in, you know, what the world thinks about Jews. So it has a particular interest there. I don't think that's a particularly Jewish problem. You know, there's lots of groups that could be interested in the same way. I'm hoping it's a reason for Jewish communities and Jewish leaders to take AI more seriously.

Yehuda Kurtzer (11:26)

Let me go back to the beginning of this question where you said the Jewish community is a little bit behind. Do you see this as a kind of common problem that Jews have historically had as it relates to technological advancement? I go back to conversation I had with miha Goodman about 10 years ago at Hartman about social media and the risks around social media. And he was probably more right than I was. But I talked a little bit about like how we are always going to be behind at least those of us who come from a kind of a traditional orientation because there's always a phase of resistance to technology and then gradually acceptance that it's taking over. I use the example of machine made matzah. Resistance, resistance, adaptation. Okay, fine, it can be kosher. That seems less damaging. Although it probably did cost some people some jobs.

David Svi Kalman (12:14)

Yeah, it cost widows and orphans specifically jobs, because they were the ones who were doing the matzo production.

Yehuda Kurtzer (12:19)

Yeah, well, that's not great. So do you think that Jews have a particular issue around this? Because of our relationship to tradition or religious communities in general are going to have the same kind of obstacles in confronting new things.

David Svi Kalman (12:32)

So Judaism, like other religious traditions, confronts new technologies all the time. We have a long track record of doing that, although it's understudied, unfortunately. Sometimes we confront a new technology, we decide to embrace it, and that turns out to be a good idea. Sometimes we reject it. And that's a good idea. I mean, like, it's kind of all over the place. And I don't think there's a kind of uniform perspective on this. I would say, though, that Judaism, like other traditions that emerged, say, before the Industrial revolution, before the 19th, 20th century, are fundamentally in a world where the understanding of history is one of more or less things don't change. We're going to establish rules that are more or less static because that's what the world needs and will change slowly. And I think Judaism, like other religious traditions, is really struggling to adapt to a world where the ethical needs of the moment are changing constantly all the time. And what ends up happening is this, I think, pretty dangerous dynamic where you have tech firms coming out with new technologies all the time with no clear ethical rules about how you're supposed to use them. The public kind of confronts them based on their existing morality. And then after the public has already figured out that they're going to use these technologies, religious institutions are like, hey, we should probably think about this. We should probably do something about this. But by the time you get to that point, your space for actually informing people's ethical decisions, their behaviors, is so small because everyone's already using it. So I think what has to happen, really, the only move you can make is to have ethicists working as fast or faster than technology companies themselves, which actually isn't impossible, really, because all the technologies that we are developing right now have been envisioned for decades. Right. Like, the idea of AI is not new. The AI at the quality that we have right now has been envisioned for a very long time. And you can imagine kind of speculating, like, what does it mean to live in a world where this technology exists? What kind of behaviors ought we to establish? I think if you want religious institutions, religious leaders, to be able to respond adequately, forcefully to ethical problems, they have to be thinking at that level. And importantly, they have to be doing something that tech companies are very good at and religious communities are really bad at, which is move fast and break things. By which I mean tech companies, especially AI, are really used to being in a world where you come up with a product, it's bad in all kinds of ways, but also good. And then you kind of make it better and better and better. I don't imagine that religious responses to AI are going to be perfect the first time around, but I think they need to be there and be there fast. And you can iterate on them based on the way people respond to them, what works, what doesn't work. But you kind of have to borrow a chapter from the playbook that tech companies are using about what it means to be effective in a fast moving world.

Yehuda Kurtzer (15:09)

It's just hard to imagine because tech companies and militaries are going to make a whole set of decisions that claim to be in relationship to ethics, but are really motivated by totally different motivations. I mean, the ethical conversation may have been coterminous with the rise of the creation of the nuclear bomb, but ultimately the decision making around producing the bomb and then using the bomb wasn't really going to take into consideration what the ethicists were arguing about, the implications of the nuclear bomb. So I think we've basically seen it borne out that the ethicists who I know, who worked at social media companies are basically not there anymore. So even if you were able to say the ethicists should be embedded in these companies, or at least working in parallel, do we have anything at our disposal that actually enables us to hold some of this technology in check when you're competing basically against capitalism and the market?

David Svi Kalman (16:07)

Not really, because the companies just move very fast. So no, I don't think it's possible once a technology has reached a level of commercial availability to really do much about it. So in some ways the AI conversation, I mean, it's worth talking about, but the scope of making change is relatively limited. I'm thinking about what do we do in terms of bioethics, in terms of genetically editing the DNA of offspring and those kind of decisions, because those are places where a technology is not yet commercially available, but it's also clearly going to become commercially available. Or, you know, Meta and other companies are working on wearables, glasses that show you computer screens in front of your eyes all the time. Again, it's not yet commercially available for most people, but the technology is getting better all the time. And you can imagine responses to those developing right now before 100 million people have them and it's kind of too late. I don't think that means being embedded in the tech companies. What it does mean is being sufficiently aware of the trajectory of the technologies themselves. The way that the tech companies are talking about, the way they want to develop it and kind of responding to that in that way. I would say there's a kind of need to be futurists along with the tech companies themselves. There has been a lot of talk recently about how tech companies kind of speak in religious language. Peter Thiel talking about the Antichrist. But it's not just Peter Thiel. There's lots of folks that use explicitly or implicitly religious terms. In thinking about what AI does for the world, religious institutions sometimes respondent enter the conversations because they're like, oh, someone mentioned my keyword. So like, I'm going to talk about it now. But part of what that means is also painting a picture of the future that is just as clear and vivid as the picture that the tech companies themselves are painting. Because if you just say, yeah, I heard what you said, and no, I don't like that, that's really not sufficient. There has to be some clear perspective of where you're trying to go, understand that the world is going to change in some way.

Yehuda Kurtzer (17:57)

I had that same reaction about the way tech companies, but also people adjacent to this conversation speak in religious language, even just by reading some of the book titles, portending what AI is going to do to us, which are just trafficking and apocalypticism, like religion invented that too. This is going to actually destroy human civilization. And that's where my own skepticism starts kicking in. I'm like, they thought we were all going to die on Y2K. Let's stay in AI for now before moving on to gene editing. So what you're asking us to do or encouraging us to do is to think in visionary ways the ways that the tech companies are thinking about where we're going to go. And let's imagine that given the pace of advancement, even since the release of ChatGPT three years ago and to now has been dramatic in this sector, where are we going to be in five years as relates to AI, that you have a clear enough vision that can say, here's what it would look like for us to not just keep pace to it, not just react and respond to it, but to develop our own thinking, to be able to either hold some of the most dangerous implications in check or to build a parallel world that could make us immune from some of the damage. This is going to do.

David Svi Kalman (19:09)

Yeah. So the tricky answer is that we kind of don't know where we're going to be even three years from now. And the range of possibilities is anywhere from more or less where we are now. The AI at the same level it is now, but maybe more integrated into companies, into society, to everyone on earth being dead, literally. That is the range that people are talking about right now. Right. There's a book that Eliezer Yudkowski came up with. I think the title of the book is like, if anyone builds this, Everyone Dies, which is a great title, but it's really that range. I think the fact that there is that lack of clarity actually works to the advantage of tech companies because it allows the rest of the world to kind of not be able to envision really well what is about to happen. But I definitely think we're talking about the loss of jobs. We're talking about the loss of especially entry level jobs, the loss of a lot of creative work, and the loss of a lot of skill sets that right now we take for granted. I like to think about AI sometimes using the metaphor of addiction and that sometimes people in the early stages of addiction, and it feels awesome, right, because you both have the benefit of your fully healthy body together with whatever substance you're kind of adding on top of that. And what happens always is the substance kind of erodes something within you because bodies are efficient that way. So right now we have a workforce that is very good at writing, an AI that is very good at writing. And it's like, great, okay, we can do this in two different ways. But inevitably one of those things is going to erode the other. Right? If people turn to AI for writing companions, especially as part of their training within the educational system, when they're in high school, in college, that is going to erode the way that people write. So part of what needs to happen I think in advance of that is thinking actively. What are the skills that we want to preserve that we didn't think we needed to talk about having to preserve because they were always going to be there. So what that means in aggregate is hard to tell. I think it's the kind of thing that communities actually need to talk about, overrun themselves. But I'm going to give you one really concrete suggestion of something that you can do to kind of push back against that. There has been a lot of talk in the last few years about AI use in sermon writing. I know there are rabbis who have talked really excitedly about using AI in sermon writing or have said here's a sermon. And then at the end they're like, surprise. This whole sermon was written by ChatGPT. What do you think about that? I think it is worth establishing a rule that says AI should not be used to write your sermons. Every word that a congregation hears should be written by a human being. Doesn't mean you can't use AI for research. I think that's fine. But the writing itself, I think should be done by humans. What's the reason for that? Part of the reason for that is it establishes a kind of public practice of here is a line against AI that we're not going to cross. Not because AI is bad at serving writing off and it's quite good, but because it gives a congregation a sense of like, this is a space for humans to interact with other humans. And yeah, it's a little bit annoying. It would be more efficient to use AI in that context in some ways. But it's one of those moves that right now feels minor. And as more and more writing tasks and public speaking tasks get taken over by AI ends up becoming a bigger and bigger deal, it's like, oh, wait, this is a space where I am assured that I am interacting with other humans who are just telling me what they think, as opposed to kind of something that is mediated by AI. I think that can be very powerful. And it seems like it's what the public actually wants. There have been surveys in the last couple months about how the public feels about AI. The public is much more negative about AI than tech developers are. And among the places where the public is least comfortable having AI being used, religion is like, they don't want AI there. So I think it already responds to something that I think a lot of religious folks are already feeling and gets to that sense of like, we need to proactively create human spaces, spaces where AI is non involved, even if efficiency would kind of push us to put AI in all these spaces instead.

Yehuda Kurtzer (22:55)

Okay, so now I want to ask you a devil's advocate question. I'll speak personally about this. I've never used AI to write anything and I've never been tempted. And the reason I'm not tempted is because, like, this is something I'm really good at and I love doing it and I love the work of thinking. And I kind of feel like I never write for the purposes of efficiency. Right. I don't write website copy. I don't write other things where like, it needs to just be done quickly and basically good enough. And I would feel really bad about myself if the thing that I do in the world, I was just taking shortcuts to do. I would feel dumb. Like, I don't, okay, you don't need me anymore. Just run this podcast with a bot and I'm with you about sermons, especially, because I feel there's a whole other layer which is I believe, for the most part, people don't go to shulem to hear a perfectly articulated sermon which summarizes a set of issues in the world. They care about the person who's giving it. So it has to be relational. But in both of those cases, it sets up a dynamic where there's something unique about the talent that's doing that writing that actually generates greater hierarchy between those who can and should be original writers and those who shouldn't. And I think one of the things that AI is actually doing, which is also a threat to the status quo, is that it is shortening the distance between great writers and good writers. In other words, it has a kind of leveling effect. So I would imagine that part of the appeal for a lot of people is I'm not going to come up with these ideas of my own, but I actually could help my company and I could advance my career if I was able to do so. So how do you take into consideration that. That, I don't know, appeal to egalitarianism is also, in some ways a moral driver of why these tools might ultimately be good for and legitimate to be used in a whole bunch of cases.

David Svi Kalman (24:57)

I think you're describing exactly the dynamics that religious institutions, religious leaders need to work through. And it's not a one size fits all thing. One of the things that's different about the AI conversation now than it was three years ago is that now the use cases are so much clearer. We can talk about, okay, AI in sermon writing. AI to replicate dead grandparents. Let's talk about each of those cases individually and think about the utility of each one. So I think it's worth doing that, but it needs to be an explicit conversation, because what happens otherwise is everyone kind of has a gut reaction that sometimes is the position you hope people get it to, and sometimes it's not. There are instances of AI use where everyone is basically on the same page of like, this is bad. So there is AI that's available that can take a picture of a person and digitally remove their clothes. And everyone's basically like, yeah, that's not good. That's dangerous. That is open to all kinds of abuse. We should regulate that. And it is being regulated and legislated in many places. But cases like writing are a little bit less clear. And often people, especially writers, have a kind of gut sense of, like, I do like this, I don't like this. Often there is a generational divide about how it should be used. I don't think those gut instincts are enough. I think there has to be an articulation of, like, what is that deeper value? Is the value in the sermon to hear a great piece of wisdom? Is it to connect to that person? And how is that different from using AI in the context of some kind of work presentation and which kind of work presentation? It actually requires that level of specificity to develop an approach, because otherwise what ends up happening is it's just the economic viability and utility that wins out at the end of the day. Those are really messy conversations. And they're conversations that happen, I think, best within the context of local communities as opposed to nationally, because you need the trust of the people around you and their instincts about what is right and wrong. And actually, religious communities conveniently can serve that role, even if they're not intended to kind of like talk through tech issues. I think actually can play an important role in helping people actually think through something that maybe was sitting in the back of their mind as a kind of anxiety, but they can now bring forward as like, okay, let's actually talk about this. What do I feel is right on this issue?

Yehuda Kurtzer (27:08)

It feels like there's another possible strategy that's out there in the world which would say, if this is going to be what people imagine as the font of at least knowledge, let's assume not wisdom, but knowledge. That the key operative strategy, both on, for instance, the anti Semitism front, but even on the ethics front, would be to feed the information that we know we need to be in the system, into the system more systematically and more effectively. Stephanie had this experience recently. You know, she teaches 8th graders and she asked, I think she asked ChatGPT, or maybe it was Claude, some version of a question of, like, how would you suggest I present the ethics of using AI and got like, an incredibly robust set of answers, like, here might be all the questions that you would introduce. And I found that a striking example because it means that there's the backing of ethicists who even unconsciously are having their material fed into the algorithms to be able to produce this information. Is that a plausible strategy on antisemitism, for instance, of like, flood the zone with the kinds of knowledge and frameworks that would tilt the predictable vocabulary that would come out on the Other end.

David Svi Kalman (28:21)

It might be that question in some ways feels more similar to questions we're already having about biases on the Internet or is Wikipedia inherently biased or things like that. And I can imagine exerted efforts to either add additional content or better content, or to make sure that the folks at the tech companies themselves are being really clear about bias, or that there's legislation around lack of bias, things like that. So I can imagine that happening. And maybe that is the right path forward.

Yehuda Kurtzer (28:50)

It could potentially work.

David Svi Kalman (28:52)

Maybe it'll work.

Yehuda Kurtzer (28:53)

It's just our limitation is that as it relates to Jews, there's just very, very few of us. And also it would essentially be a means of capitulation. Right. Like there's no way to actually fight this on the other end. Because ultimately it feels like the questions of ethics are coming down to kind of two choices. You can either have questions around should this technology exist or once this technology exists, how do we engage with it? Or how do we influence it to be on our side? So that would be a straight capitulation to the second set of questions. I can't really fight that this is going to take over how knowledge is produced in the world. I'm simply going to try to rig the outcome of what knowledge should look like on the other end.

David Svi Kalman (29:30)

Yeah, I mean, look, there are folks already who are writing with the explicit understanding that they are not writing for humans, they are writing for AI. I think like Tyler Cowen wrote about his most recent book that like, you know, it is a book for the algorithms, it's not a book for human beings. And there are folks who will encourage you, like, yes, if you want your ideas to live on in prosperity, write for the Internet. Some folks have kind of done this unintentionally. The website halachapedia, which is a kind of compendium of ideas about halacha, it was really just intended for folks who are interested and maybe aren't specialists to learn about Jewish law is now a kind of central resource for AI. When people ask it halachic questions, which I don't think was intended, but I can imagine all kinds of resources being developed that have that in mind. What is tough is that the Internet's a pretty anti Semitic place. And if it's a question of the raw balance of anti Semitic content versus non anti Semitic content, I'm not sure that it's possible to produce enough non anticipated content to actually just in sheer numbers, counterbalance that. I think I remember reading that one of the reasons that Mecca Hitler became Mecca Hitler, like One of the reasons that Grok kind of turned towards the Nazis was a little tweak in the code that was basically like, take the Internet a little bit more literally, which turns out to be very dangerous. So it might require something more subtle than just a kind of influx of new data.

Yehuda Kurtzer (30:50)

Your description of the Internet is fundamentally anti Semitic calls to mind the incredible song from Wreck It Ralph sung by Sarah Silverman, which she says this ode to what she calls a place called Slaughter Race. And it's basically like a mocking of like kind of Grand Theft Auto. But it actually is kind of what the Internet is. It was like this beautiful ode to. That's what actually takes place here. So by the same token, trying to actually enter into those spaces feels like you have to just suit up and accept that you're putting yourself in a little bit of a cesspool, I'll say.

David Svi Kalman (31:18)

One thing that makes me hopeful is that there is some evidence that unwanted behaviors in AI are linked to each other. So there's this concept in AI called alignment, right? That your goals and AI's goals are kind of the same. There is research done, I want to say, around a year ago, suggesting that it is possible to intentionally misalign AIs by providing a kind of like, bad data, negative data, and that once an AI is misaligned, it ends up being misaligned on a whole host of paths. So it gives you bad code and it's misogynistic and it's racist and it's anti Semitic.

Yehuda Kurtzer (31:52)

So intersectionality, right?

David Svi Kalman (31:55)

I mean, like, in some ways it makes me hopeful in that it suggests that there is this idea of the good that we are kind of trying to get AI to move towards, and that antisemitism is something that will be pushed against in aggregate, even if the folks designing it are not specifically interested in anti Semitism.

Yehuda Kurtzer (32:15)

But this opens up something actually very much deeper which would suggest that if part of the goal, even on the AI side, is alignment, then a person who was trying to gather knowledge or produce something would want to basically disclose their biases at the outset, and then AI would work to advance their knowledge in service of those biases. If you said, these are my political commitments, these are my moral commitments, help me think through this question, then you're going to lean on the good inclination of the AI tools as opposed to, I'm looking for information that explains why Jews drink the blood of Christian babies, where it's just going to provide you the information that you need. And I think the thing that feels deeper about that is that one of the things I think is that the crux of this issue that religion has a hard time with is that these are ultimately tools that are designed for human beings in a deeply atomized way, and religion really does not know what to do with that. Right. So just to make it clear, you gave the Halachaedia example I was preparing. I was thinking back to the academic work of one of our former colleagues, Yakir Englander, who studied the phenomenon of young ultra Orthodox men searching for response on the Internet, mostly connected to sexual behavior, because they were embarrassed to ask those questions. So they would go online and there were rabbis who were kind of giving responsa online, and then they got aggregated into websites that you could actually kind of choose your own responsum, which is totally wacky and totally not what responsa was supposed to be about. I have a relationship with a rabbi. They give me a customized answer. Instead, I'm actually shopping for this piece of information that I need to use in order to figure out what I want to be religiously. It feels to me like that's the threat to religion is that we lose any coherent notion of authority that's invested in particular individuals and some kind of collective shared knowledge that all of us subscribe to that I can have a slightly different opinion within it. And instead any single individual can kind of figure out what Judaism is or what they want Judaism to say by means of these technologies. That feels like completely unwinnable.

David Svi Kalman (34:23)

Yeah, to some degree this feels like an extension of something that has been happening for the last 50, 60 years where there has been a huge push to make Jewish knowledge more accessible to folks who are not in a position to decide Jewish law. And because of that, the texts that would previously have been used by rabbis to say, like, oh, this is a person who you should take seriously, and their position you should take seriously. Now, anybody can use the whole genre of the halachic essay written by a non rabbi or written by a rabbi, but not for halachic use, has been around for a very long time, often trying to address questions of technology without any clear sense of who's supposed to make that decision. And AI kind of like pushes that up a notch because it now means you don't need to know anything, you don't need to know Hebrew, you don't need to know anything about Jewish law, and you can get a great, very learned answer that kind of goes along with what you want. So, yeah, it does have a negative impact on authority structures. One thing that I worry would happen or will happen as a result of that, is it kind of pushes Judaism towards increasing conservatism or kind of increasing status quo. Because if no one is actually trusted to make the argument, to revise something, because you're like, anybody can make their argument. Anybody can make an argument that says anything they want. All you have left is the rules already at your disposal, and you don't really trust anybody to make any kind of argument anymore. So, yeah, I could see that as being very dangerous. I suspect in the long run, it will mean that the way that we think about Jewish law, the genres that we use will have to change to kind of accommodate that and to acknowledge the fact that, you know, there's nothing so special anymore about being able to write a responsum or being able to create a midrash or anything like that. At the same time, maybe the way that AI gets used in religious contexts and maybe the standards that are set up around here are places where we don't think it's useful. Here are places where we think it would be dangerous to use, can kind of like push back against that. To say, like, yeah, you could use AI for this context, but maybe you shouldn't, because it has this larger detrimental effect on Jewish community.

Yehuda Kurtzer (36:20)

So let's go back in time a little bit. I remember when the first piece of technology that kind of democratized Jewish knowledge, at least in my lifetime, not the Encyclopedia Judaic, obviously, the floppy disks that constituted the Bar Ilan Responsa project, which became CDs, and then thumb drives. And it was like, okay, you can have basically all of Jewish texts that they were able to put on those floppy drives or the thumb drives. Disk on key, disc on key. That's right. And it was very clear for a while that was a tool that benefited scholars more than it did laypeople, because people kind of knew where they wanted to look. But you could kind of see the ways that it watered down a lot of teaching, because you could tell when someone had just typed a keyword into Bari Law. And it was like, now I'm teaching a class on this topic. And they didn't even change the font of the printout of the sources that came there. The phase two of that was Safaria, right? The massive digitization and make publicly available Jewish texts that are out there. I don't know. I find that to be like, one of the most incredible technological accomplishments that we've seen as Jews. And it's ripe for the same screwing up of authority, weakening of creativity. Right. Because now you can just, you know, copy anyone else's source sheet, or you can just get all the sources that you need. But it doesn't help you think about how to make an argument. Doesn't require you to know the material in the original. Even I say these words and I sound like a hopeless elitist. Where's the boundary, then, where you would say, when it comes to these technological advancements, when it comes to Jewish texts, between the stuff that's pretty good and the stuff that's dangerously bad?

David Svi Kalman (38:00)

Yeah. So to some degree, there's like fused AI parlance. There's a little bit of a moat still around folks who actually do know what they're talking about. One of the known issues with using AI is that you can't easily tell when the AI is saying something crazy, when you are beyond your own personal competency. Right. There's all these folks who believe that they've made breakthroughs in theoretical physics because they don't actually know anything about physics, and they're just talking to AIs, and you're like, wow, I have to tell someone about this. I think something similar happens with Jewish law, where you can get a lot, but it's not everything. In some ways, I find the question a little bit frustrating because for a lot of technologies, there is this instinct to think about, well, what is the interface between that technology and Halachah, specifically? Because it's practical, because it's the training that so many Jews who go to Jewish day schools have. And halachah is important. But at the same time, I worry that it's not the most useful frame for thinking about AI. There are more useful questions around AI in terms of humanity. What does it mean to be created in the image of God? What does it mean to value other human beings? That matters to me a lot more. The halacha piece is important. And, yeah, I can imagine people kind of coming up with crazy halachic ideas as a result of AI, But I'm less concerned about it because I want to see the other ways that Judaism can have an impact. And those are places where you're not just trying to get an answer. You're trying to come up with a new idea or a new image where I don't see AI kind of slotting so easily into something that rabbis already do. There's something more complex going on here.

Yehuda Kurtzer (39:31)

The nature of this business that we're in or that we're adjacent to Jewish leadership, Rabbinic leadership, Jewish education, all of these things is that, by and large, some people are going to be interested in the larger philosophical questions and everybody's going to be concerned about whether we're actually all going to die. That's legitimate for source of concern. No one has yet explained to me exactly how that happens, but I'm sure that there's a plan.

David Svi Kalman (39:54)

I think the folks who have it in mind are intentionally not talking about it because they know the AIs are listening.

Yehuda Kurtzer (39:59)

Really? Yeah. Okay.

David Svi Kalman (40:02)

Yeah. A separate conversation.

Yehuda Kurtzer (40:06)

I think that we all crave basically some usable lane to think about how this implicates our lives and our choices. Like, forget about halacha for a second. Think of in the field of Jewish education. One of the things driving questions right now in Jewish education is how do you make this more affordable? People who want it are stuck. The schools can't really market themselves to people who aren't coming in yet because they don't have the value proposition yet. To be able to say, this is worth you're spending $40,000 a year on it. You're going to see a gravitation towards AI technologies as like, oh, this is a really easy way for us to dramatically reduce costs if we can get all of these needs taken care of, if teachers can manage multiple classes through the use of these technologies. So I'm just using that as an example of like, yes, I want to talk to you about the human condition. But that's not where the conversation is going to live for the vast majority of institutions. Rabbis, the well intentioned ones, are going to say, I want to use this not because it's going to write a better sermon than me, but because I could actually visit more sick people. I could do all the IRreplaceable things that AI won't be able to do yet until it has robots that I could actually uniquely contribute. And so I think that's why we keep kind of gravitating towards what you called like halacha questions, like questions of use and questions of policy. And in that sense I'm kind of curious, like if that's the playing field, do you have a sense already of recommendations that you give to either educational institutions? You gave one. Rabbis should commit not to do this, but assuming that all of our fields are going to use AI one way or the other, what's the guidance that you offer to them in that space?

David Svi Kalman (41:49)

Yeah, I'll say two things. First, on where the focus is going to be. This is a constant problem with new technologies as well. One of my favorite response of all time is a rabbi in Cairo who's been being asked Can I use electricity on Shabbat? Can I use electricity on holidays? And he gives an answer. But before he answers, he says, you should know that I don't really have time to answer this question because I have a job. And these are hard questions. So the fact that this lies outside the main responsibilities of most Jewish educators is both totally understandable and also part of the reason why religious institutions struggle to provide the kind of larger ethical responses in terms of the question of, like, where this actually fits in. There's all kinds of efforts right now to see how AI slots into education and whether AI can replace teachers in whole or in part. I think the areas where this may be most successful are in mathematics and in language, where there's kind of like this rote thing, and also where if the AI can, like, gauge really, really well exactly what the student needs or, like, the right metaphor or the right exercises that can get them to the next point maybe can make that more efficient. Jewish education is both blessed and also burdened by this very heavy requirement to learn Hebrew, different modes of Hebrew. And I can imagine AI being used in some ways to speed up that process. At the same time, there is a kind of advantage to Jewish education in relationship to AI that is different from other kinds of learning. So, like, if you're thinking about prose writing, right, the ultimate point of prose writing is to make money. For most people, the ultimate point of education is to have some kind of economic benefit, right? For most people, that's why they go into the world. That's why they go to college, do those things. Now, sometimes it happens more inefficiently or more roundabout ways, but they're trying to get economic benefit. And so the AI's utility is clear. The point of learning Torah is not to make money. It's supposed to be an end in itself. It's supposed to be something that is enjoyable and absorbing and deep in itself. And so you can imagine AI kind of entering that not as a way of kind of like doing the Torah learning more efficiently, like, I get through the track day twice as fast as I would otherwise, but as a way to make a learning experience that feels more enriching and more enjoyable and more educational. I don't know exactly what that looks like. Maybe it means like creating digital avatars of maimonides and, you know, other rabbis. Maybe it means, you know, just having some kind of interface where you can kind of, like, quickly ask questions so you don't get stuck in the same ways that you would otherwise. That's the area where I'm most excited about. About seeing AI develop. Yeah, I'm sure we'll see tons of experiments about this stuff in the next few years.

Yehuda Kurtzer (44:27)

I'm going to ask you two last questions. One more conceptual and then one more practical. The conceptual one is you wrote an essay, a kind of, I think, the first major piece of Jewish thought, trying to kind of aggregate the state of the field a little bit. It's on your website, and one of the things you said at the outset of the essay was it's hard to tell whether this is a case where it's like, okay, I know exactly what Jewish texts I'm bringing to bear. Right. To be able to determine this, or whether we're confronting something that's totally new and that whenever Jewish thinkers confront technology, there's a tell a little bit by whichever text they use right away, because it's kind of disclosing what they think of the issue. Right? Text by analogy, are there obvious texts that you're going to these days in thinking about AI that you would prescribe as a kind of Jewish study curriculum for people who are interested in thinking about this question?

David Svi Kalman (45:19)

So I would say, speaking really, really abstractly, really, really broadly, that there is no Jewish perspective on AI. It does not exist as such. And I think it's important to say that, because otherwise what ends up happening is you end up in this endless search for, like, let me find the text that is most similar to what ChatGPT put out yesterday. And that both takes a lot of time and also takes energy away from the work right now to think about what behaviors and practices are we trying to develop. It's not that there's nothing that's relevant, but I think the basic assumption should be there's not going to be so much that you're going to find. The work is on you. So to the degree that a Jewish approach to AI develops, it's an approach that develops now that people are developing in practice. So in 50 years, maybe we can say, like, oh, we developed a Jewish approach to AI. And that was amazing, and that was really important. I hope we end up in that place. But I don't think it's quite there yet. In terms of the text that I do find useful for thinking about this. The one place where I think, or I'm noticing that Judaism seems to have something specific to say is on the question of AI's moral status and relationship to human beings, I see a lot of Christian denominations heading in the direction of we care about biological humanity and not behavioral humanity. Right. It matters if you have a physical human heart. It doesn't matter if you talk like a person. If you do not have a physical human heart, then I could treat you like whatever. It doesn't matter. It doesn't matter how you act. I would say that Jewish sources tend in a different direction. There seems to be an interest in deciding human behavior on the basis of how does a being act, a being that seems human sometimes get the status of a human, or a demon that acts human is treated to some degree like a human. So I think those texts are helpful in developing some kind of Jewish approach. The other text that I think I've actually quoted on this podcast before is from the brother of the Maharal of Prague, who's one of the earliest opponents to the printing press, or at least for Jewish context. And one of the arguments he makes for why you shouldn't use the printing press is he says, imagine a beggar who kind of comes to town and borrows a bunch of stuff from a bunch of different people and ends up being dressed in really nice clothes as a result. Second beggar comes to town, sees the first beggar, and it's like, oh, that guy is really wealthy, and goes to that first beggar and is like, can you give me some money? Can you give me some clothes? Not realizing that everything the first beggar has is actually just borrowed. So this is his way of talking about how a book, a physical book, is basically a kind of borrowing of the real wisdom, which is the wisdom of people. And if you learn from a book, then you are being deceived in some way because you are only getting a borrowing. You are borrowing from something that itself is borrowed. You can imagine AI kind of functioning similarly in that it gives you this impression of great depth, but that depth isn't always there. And that at the end of the day, you actually do need to go back to people, to human beings.

Yehuda Kurtzer (48:18)

Last question, which is very practical, maybe annoyingly so. And I'm sure you get asked this a lot if you had three key suggestions to Jewish institutions that are trying to formulate an AI policy. I know that usually those questions are about proprietary work, right? The ethics of copyright law, things like that, and also probably wind up getting reduced to, are we going to go with the pro suite of this company or that company? But if you were going to say, here are three Jewish principles that you would want to see Jewish institutions invest in or advance as part of their AI policies. A synagogue, a school, the Shalom Hartman Institute, what might those be?

David Svi Kalman (48:56)

I'm going to give you two. I'm not sure I can come up with three right now.

Yehuda Kurtzer (48:58)

That's two we'll get AI to do.

David Svi Kalman (49:00)

The third great one is value humans first make sure that whatever you are doing does not result in in the loss of human dignity, which can happen in really subtle ways. Because when you have AIs enter a system and do things that humans previously did, there is a way in which the AI does make people feel like they are less than adequate. And I don't think people should feel like that, even if sometimes they do need to change the work that they are doing as a result. That's one. So I would say the second one is have the conversation. A lot of people have come to AI with very different understandings of what should and should not be allowed. I think they often only realize in the room with each other how far apart they are on both the way they use AI and their assumptions about how it should be used and their own personal competence. I think there is sometimes a sense that it's not worth having those conversations because people are just going to do whatever they do anyways and AI companies are going to win at the end of the day. But it does actually make sense to have conversations internally about what do you feel comfortable with, what do you not feel comfortable with? Because those conversations need to happen somewhere. And I think there's always a sense that like, oh, someone else is doing that hard work, but the work happens best when it happens locally. And coming up with policies that you and the people you work with together feel good about, I think is probably the best way to go. It's also a way of elevating folks who are younger in the organization who often both have a clear experience with AI, use it more frequently, and also are more likely to be injured by AI development in terms of their own career prospects as they move up the.

Yehuda Kurtzer (50:39)

Career ladder has the added advantage of doing exactly that same thing that you described about doing of elevating or valuing humans in the processes by engaging them in exactly the participatory process of figuring out where do they want their own work work to begin and where do they want it to end. Daviti, thanks for being on the show today. Great to be here.

Podcast Announcer (51:06)

Here are some other things that are happening at the Shalom Hartman Institute this week. Are you a college student or the parent of a college student? The Edward Fine Winter Student Seminar is taking place in New Orleans from December 22nd to 25th. Through dialogue, text, study and reflection, college students from across North America will grapple with questions of Jewish ethics, civic responsibility, free expression, and the balance between Jewish safety and power. Together, they'll imagine a more inclusive, democratic future for both Israel and North America, a future that protects all citizens, including Jews. Registration closes October 28th. For more details and to register, click on the link in the show notes. The signing of an agreement to end the war in Gaza and to return the remaining hostages leads us to cry, to rise together, and to hope. Check out our Instagram for resources from our Israeli colleagues. For this moment, our ritual center in Jerusalem has written a prayer for the physical and spiritual recovery of the hostages, while Hana Pinchasi shares a reflection on what comes next for Israelis as the world war ends and they are fully able to mourn. Thanks for listening to our show and special thanks to our guest, David Svi Kalman. Identity Crisis is produced by me, Tessa Zitter and our executive producer is Maital Friedman. This episode was produced with assistance from Annie Beyer Chaffetz, researched by Gabrielle Feinstone and edited by Josh Allen with music provided by so called transcripts of our show are now available on our website. Typically a week after an episode airs, we're always looking for ideas for what we should cover in future episodes. So if you have a topic you'd like to hear about or if you have comments about this episode, please write to us@identitycrisisalumhartman.org for more ideas from the Shalom Hartman Institute about what's unfolding right now. Sign up for our newsletter in the show notes and subscribe to this podcast everywhere. Podcasts are available. See you next time and thanks for listening.