The following is a transcript of Episode 124 of the Identity/Crisis Podcast. Note: This is a lightly edited transcript of a conversation, please excuse any errors.
David Zvi: Hello and welcome to Identity Crisis, a podcast about news and ideas from the Shalom Hartman Institute of North America. My name is David Zvi Kalman, and I’m a scholar and resident at the Institute, and I’m filling in this week for Yehuda Kurtzer. Today is Wednesday, January 25th. 2023.
So I wanna start off today by telling you about an experiment. Back in 2016, I began reading these reports about these pieces of software called neural networks, which were basically a type of artificial intelligence or AI, that were capable of taking in huge amounts of data and spitting out new data written in a similar style.
The neural networks of the time were relatively primitive and they were hard to use, but they were intriguing. So I did what any self-respecting Jewish nerd would do. I fed a translation of the Hebrew Bible into one of them, set some parameters, and tried to get the network to write more in the same style. I’d say the results were pretty mixed. The program spit out things that looked like Bible verses, but they didn’t really stand at the scrutiny.
Sure, there were phrases like the Lord spoke to Moses and lots of references to the children of Israel, and plenty of this guy begat that guy. But the verses themselves didn’t really make any sense. The exercise was fun, but it didn’t really mean anything. I think it’s fair to say that AI has improved quickly since then.
In fact, the experience of watching AI develop over the last decade has really meant being repeatedly startled, as computers over and over again, began performing tasks that we had recently believed would remain firmly in the domain of human beings. Just three years after my little experiment, a computer model trained on images of human faces began spitting out 40 realistic images of people who never existed. If you’re curious, you can find these images that this person does not exist dot com.
In the same year, an AI language model called GPT3 became so sophisticated that I began slipping texts written by it into my classes about artificial intelligence as a kind of meta text, as a way of letting learners consider how it might feel when human beings lost their monopoly on the use of religious language.
But despite these astounding developments, the people who were most interested in AI tended to be the people interested in tech generally. Despite the fact that sophisticated AIs were already being deployed by governments and corporations, the public mostly treated them as impressive curiosities.
But something seems to have shifted, and it shifted because of a new, publicly accessible AI called ChatGPT which became available this past. If you’re not familiar with ChatGPT, let me give you a sense of its capabilities.
In the week since it’s public debut, I have seen people ask ChatGPT to write complete computer programs, compose resumes, summarize lectures, draft grant applications, teach them how to use Excel outline presentations, write sonnets, and write instructions for how to get peanut butter out of a VCR in the style of the King James Bible.
None of this I should emphasize, requires knowing any code. You type your request into a window, press a button, and watch the answer appear one word at a time in a manner of seconds. It’s clear that the creators of ChatGPT expected this AI to be disruptive to any number of fields, but there is one field that is already feeling its effects acutely, and that is the field of education, which ChatGPT seems to have turned on its head almost overnight, to the point where New York City schools have actually banned access to the site. If you’re a high school or college teacher, many of your standard writing assignments can probably now be done by a machine, making cheating much harder to detect and calling into question whether it makes sense to continue administrating those assignments at all. An article in the Atlantic went so far as to declare that college English essay was dead.
The effect on teachers is also, I think, the reason that AI has now finally become interesting to Jewish leaders, many of whom are either teachers themselves or run in the same circles as teachers. Given all of this, the topic for today’s show is what do ChatGPT and other AIs mean for Jewish education? What do they mean for Jewish learning? And what do they mean for how we think about texts as sacred?
Now, I’ll confess that I have been playing with ChatGPT too, and when I asked it to suggest topics for future episodes of Identity Crisis, its first suggestion was, and I quote, a podcast that explores contemporary issues of Jewish ethics and moral reasoning. When I then asked it who an appropriate guest might be, it responded, a senior faculty member or scholar at the Shalom Hartman Institute who could provide expert insight and analysis on specific topics. Neither of these things is exactly earth-shattering, but they do happen to fit the description of today’s topic and today’s guest, who is Rabbanit Sara Wolkenfeld.
Sarah is Chief Learning Officer at Sefaria, a new online database and interface for Jewish texts. She’s also a rabbinic fellow of the David Hartman Center. Sarah, welcome to the show.
Sara: Thank you so much, David Zvi, it’s great to be here.
David Zvi: So I, I want to talk about AI and ChatGPT but before we do, I wanna start out by talking a little bit more generally about the relationship between Jewish education and software development. Sefaria, which I know you call new in your bio, but in fact just celebrated its 10th anniversary, has become the go-to site for basically anyone interested in free access to Judaism’s core texts, many of whom you’ve made available in English, including a full translation, a full free translation of the um Babylonian and Palestinian Talmuds.
What does the relationship between pedagogy and coding look like from your perspective? What does it feel like to sit somewhere between Torah and Silicon Valley and to be working with people who are kind of have a foot in both? So first of all, just say David Zvi, apropos your introduction, you raise the question of what makes text sacred. And I will say that many, most, the vast majority of Judaism’s sacred texts, what we would call sacred texts, are very, very old. So from that perspective, I would say Sefaria is still very, very new and so I have a right to use that in my bio.
In terms of your question, it’s obviously, or obvious to me that it’s really exciting to be at this nexus of education, Jewish education, pedagogy, and coding developments in new technology. I believe that Jewish texts were always meant to push us towards innovation. I think that we have a very long tradition of seeing our traditional texts as something that would help us navigate cutting-edge questions in the modern world.
I think that if we didn’t believe that as a community, things would very quickly stagnate, and so I think it’s appropriate that now new developments in technology both challenge our ideas about sacred text and also allow us to bring that forward in new ways.
I sometimes say to teachers that it’s very clear to me on a kind of individual basis when I talk to teachers one-on-one or in small groups, that they think they’re doing a good job and the best job they possibly can and I respect that. And it’s also very clear to me when I talk to Jewish educators, especially large groups of Jewish educators, that no one feels that Jewish education in, let’s say the Western world in 2023, is perfect or amazing or without room for improvement.
And so I think that balancing those two pieces are of healthy respect for the practices that we’ve had for a long time, but also a real openness to the changes and the transformations that can be brought by new technology, that nexus is exactly where I sit. And it’s a great place to be because I really believe in the power of Jewish texts to be transformative. I’d say I’m a newer convert to the idea that coding can be transformative. I don’t have a background in information technology, in computer science, but it’s been a really amazing place to situate myself.
Very early on in my time in Sefaria, when I began working with engineers, had a conversation with Brett Lockspeiser, with one of Sefaria’s co-founders, about this idea of how we could improve education and what we could change. Really at that point, it was about the Jewish studies classroom. And I talked about questions that educators raised and questions that educators felt were maybe unanswerable, and he said to me that from an engineering perspective, there’s not really an idea of an unanswerable question. That all the questions I would bring to him, the question was just how easy or how hard is this to answer?
That shift in perspective has really guided my work at Sefaria. It doesn’t mean that it’s limitless, what we can do with digital technology. It just means that it’s a different perspective, a different approach to thinking about these questions that really constantly opens up new possibilities, which again, I believe is the job of Torah.
David Zvi: So that, that’s really interesting. And I wanted to pick up on something you said, around a kind of attitude of openness towards new technologies. There’s a couple ways of interpreting that. One is that basically Torah exists in whatever format it exists, and it can exist in scrolls, it can exist in books that can exist online, whatever.
The other one is to imagine that there is something deeper or more permanent that is happening in the relatively new interaction between Jewish texts and online environments that is actually irreversible, that has actually done something which now makes the study of Jewish texts in some way wrapped up in online resources, in databases, in the ability to search for keywords, things like that. How do you see that connection?
Sara: I once spoke to a rabbi of a large synagogue about the question of using Sefaria to teach his classes, and he said to me, we really need to study from books. It’s really important that my congregation used books to learn Torah the way that Jews have been doing for thousands and thousands of years.
And I thought to myself,
David Zvi: That’s like a layup for you.
Sara: I, I, I didn’t, I actually did. I did not react strongly in that conversation. I thanked him and moved on, but I thought to myself, well, I’m not a historian, but I don’t think that that is a true story. And in fact, I’m quite confident that that’s not a true story.
And so, in answer to your question, I believe that it does make a difference in what format we study Jewish text. So in that sense, I’m a, I’m a believer in the strong version. I think it definitely does make a difference whether you encounter texts on an app, let’s say on your phone, on a screen, in a book, in a scroll.
The main way in which I think this makes a difference is that I think that different technologies are good for different things, and I think the, maybe the best proof of this is the fact that Jews worldwide still use scrolls. And David Zvi, maybe you would know the answer to this question better than I do, but I think we might be the only ones. Like, I don’t know other stories of communities using scrolls on a large scale anymore. And I think if we do that, it’s because we’ve decided that that technology is really good for certain things.
We could talk about what those things are. It’s really impressive. It preserves certain features that we think are important to the experience of reading Torah, namely memorization, and the idea that oral and written Torah are sort of essentially interwoven in the experience of reading from the Torah in the sense that it’s not vocalized, it doesn’t have cancellation marks in the text. There’s something really powerful about that.
Ultimately, I believe that our community will hold on to the technologies that are powerful for particular reasons, and we’ll let go of the ones that are not powerful. I don’t just believe that about the Jewish community. I think that that’s just true about technological progress in general. So I think it does make a difference. I think that our challenge sitting in this place of an abundance of options is to be really thoughtful about the technological choices that we’re making.
So sometimes teachers will tell me, I really want students to stay in the book. I don’t want them on a screen. And I think that’s fine. And I tell them that. And I just wanna know what is it about the book that you think is great? And is there anything that you feel your students are not getting from a book that you might be able to get from a different technology?
When we frame it all in terms of different technologies, when we help educators, students, I think most importantly, students realize that books are also a technology, and pen and paper are also a technology, then I think we can open up those options. So going back to that openness that you talked about before, and really think through what’s going to help us achieve our goals for the study of Torah, which by the way, different people have different goals for that study.
David Zvi: That’s really useful. And I think it’s, there’s something poetic about thinking about Torah as, as a kind of study or Jewish text is a kind of practice that spans literally the entire range of technologies for transferring media from the scroll to this is the computer. A
t the same time, and I think you’ve written about this actually, those transitions from scroll to book, you know, from oral to written aren’t, not peaceful. There are quite a few, uh, debates, arguments, real anxieties that come out, especially from rabbis in those moments of transition.
And I’m curious whether you feel like there are lessons to be learned from either the shift from oral to written Torah, or from manuscripts to printed books that are useful for this moment we are encountering now, we have an AI that is able to replicate Jewish texts, to kind of source ideas or proof texts almost from nothing.
Sara: I do think it’s very important to look back at those past media shifts. I think it’s maybe especially instructive, maybe because we’re so far removed from it to look back at the way the Talmud describes the shift from oral to written Torah, and I wanna be clear that I am not necessarily talking about what happened historically. Historically, scholars debate at what point things actually got written down and seems pretty clear that by the time that the Babylonian Talmud is intensely debating whether or not it’s okay to write down oral Torah, that ship had already sailed.
But from a literary perspective, in terms of the way that the Talmud constructs that conversation, I think that there is a lot to learn. I think you see there, again, a tremendous openness to really thinking carefully about, what does this new technology offer us, and not just responding a, with fear, what could this possibly do? Couldn’t this destroy our community? Or b, with a sense of entrenched traditionalism. No, this is the way we’ve always done it. And so it can’t possibly be done any other way.
And the other thing that I really like about that conversation that is a little distinct from the conversation that happens when you think about manuscripts to printing press, is that God is a player in that conversation. In the Talmudic imagination, Talmudic composition of what’s happening there, God really wanted there to be an oral Torah, and so we’re meant to value that in some way, and we’re meant to be careful and approach with some trepidation the idea of writing things down, because there are downsides to written Torah, just like there are downsides to oral Torah, just like there are downsides to every technology and also upsides.
And so for the rabbis to decide that ultimately, yeah, the pros really do outweigh the cons there. It’s a really bold and audacious move, and I think it should help us open up our minds to thinking about what can ChatGPT, for example, offer us as educators. It’s easy, it’s much easier to think about, oh my goodness, how could this bring down the whole system?
And there’s a fear there, and that really has to be grappled with. But I think it’s more exciting and more interesting and more authentic to the text of our tradition, to think about what can we do with this? How can we move tur forward using this new technology?
David Zvi: So I I, I like the balance that you’re setting up. On the other hand, like I’m a little, I wanna be a little bit of a realist for about this for a second. So there’s a passage that I love from Chaim Ben Betzalel, who’s a 16th-century rabbi, the brother of Maharal, who did make a golem, and he is a strong critic of the printing press. One of the strongest critics I think of the use of the printing press for this dissemination of Jewish knowledge.
And he introduces this analogy in this screed against the printing press where he says, the printing press is like, imagine a poor man who has few clothes and goes around to people who are wealthier than him borrowing clothes from other people. And so there’s a poor man who is now dressed in a rich person’s clothes, and now another poor man comes up to this poor man dressed in rich person’s clothes saying, can I borrow clothes from you? Not realizing that the clothes that this poor person is wearing are not some signal of an underlying wealth, but they’re simply being borrowed from other sources.
So Chaim Ben Betzalel’s idea is that the printing press is this kind of borrowing, it’s, it’s a borrowing of these founds of knowledge, these kinds of sources of knowledge that are located in actual human beings. And when you study from books, you end up with this false sense that you are gaining this deep knowledge when in fact you are just looking at something that is borrowed from these other sources.
So, so like I both like the idea. And then also I’m cognizant of the fact that like, he lost, like he totally lost. And like you can talk about the tension, but at the end of the day, a lot of the time what happens is the new technology just wins. And actually the example you raised before about the Torah scroll is both wonderful and also deeply frustrating because we actually have no idea why Jews kept using scrolls.
Basically Christians very early on started using codices, and Jews just didn’t, and actually it’s Christians who are the first to remark on the fact the Jews are these Torah scroll people long before the Jews own it as their own, at least from the evidence that we have so far. So I both love the model of it, but I’m also wary about how it plays out in practice.
Sara: There’s a lot to unpack there. I’m tempted to go back to the scrolls, but I’ll, I’ll stick with your, the first part of your question first. I think that it’s not just a balancing act in, sort of, I don’t just wanna pay lip service to a balancing act. I think that it has to be deeply integrated into our pedagogy.
So I hear the point. Sure, when you read a book, you’re seeing knowledge that is in some way not real, because, kind of a way of saying that is like, anyone could write a book or I could take something that I learned from you and write it into a book and it’s not really about my knowledge. And I hear that at the same time, my very first instinctive response to the story is, wow, how empowering for that poor person, how empowering to go around with other people thinking and treating you as though you have something to give.
And I think that that really plays out as an analogy to students, to really students of any age who are learning a Torah. I think that the whole model of having students, the whole idea of Jewish education can only work in the long term if we believe that we’re empowering people to pass that on to others.
And if we’re gonna say, no, no, no. It was only like the top 10. Call it the 1%, only the 1% get to teach other people. Well, that’s not a lot of people. There’s not that many Jews in the world. I don’t think that’s a good model. I don’t think that’s a good idea for us to pursue as a community. I think for this to work, we have to assume that as many people as possible should be disseminating Torah. And that means that, yeah, they’re gonna be the poor people in the rich person’s clothing sometimes, but it’s not fake because that poor person actually has some clothing on their body and maybe they can share it in some way.
One of my favorite pieces of Gemara is this story about these two rabbis who were sort of in competition with each other for who’s the better rabbi, really. One is the better rabbi because he remembers everything about Torah and Torah can be reconstructed solely by him. But Rabbi Chiya is the flip side of that. And he says, well, the reason I’m the better teacher is because actually I, I grow flax and we weave them and we create nets and we catch animals, and then we have hides and we can write Torah scrolls, and then I give out those Torah scrolls to all my students and they teach them to other people who teach them to other people, who teach them to other people.
I think technologically in that story, there’s a lot going on where Rabbi Chiya is basically saying is like, yeah, when you hold all that knowledge, when you’re the 1%, it’s just you. It’s just you and whoever’s in the room with you. Especially at a time when there wasn’t Zoom and there weren’t technologies for getting your word out beyond the people in the room and Rabbi Chiya is saying, no, like that’s not an efficient model in the long run. Actually, we do need to empower people with technologies that go further and are those people, quote unquote, the real deal? Well, not in the same way as Rabbi Chiya is, but that doesn’t mean that they don’t have something to offer.
So I agree with sort of going into it with eyes wide open. I agree about this balance piece, and I also think, again, that if we’re only gonna hold onto those people who really hold all the knowledge within themselves, those are the only disseminators of our tradition, then we’re going to lose out in a big way.
David Zvi: I, I’m reminded in, in hearing what you’re saying about the Lubavitcher Rebbe who was a huge proponent of the dissemination of Torah widely, saying, if, you know aleph, teach aleph. That there’s no shame in knowing less Torah than the next person. That everyone is potentially a source of knowledge. So I think there’s something very nice about that, and I think it nicely paints a picture of this set of revolutions around the dissemination of Jewish ideas as one that is more or less about people in authority losing power and that power being distributed more evenly.
There’s something very appealing about that. But I’m curious, and here I wanna turn to ChatGPT, and exactly what the limits oof that are. Last night I was playing around with ChatGPT and I asked it to write a sermon for me about why New York bagels are better than Montreal Bagels, which should say, I do not endorse that, I think Montreal Bagels are better, but whatever. And it said a bunch of things.
One of the things it said is, it quoted the phrase, “maaseh avot siman l’banim,” that the deeds of the fathers are a sign for the children, to reinforce the idea that one of the things that makes New York Bagels better is that they are this traditional product that the way that they’re produced is passed down from, you know, from parents to children and their family businesses and all the rest. There’s something kind of nice about that.
So, so on the one hand, I love this idea of this distribution of knowledge of making ideas radically more accessible than they would’ve been otherwise. And I should say on that, part of the reason that we even have responses already, a corpus of Jewish thought about artificial intelligence is because of that distribution of knowledge. Most of that is produced by people who actually have no decision-making authority, who are just putting out ideas because they have the knowledge and they’re able to do so. So there’s like something wonderful about that.
On the other hand, I also believe that there is some importance to the idea of Torah ease, for lack of a better word, of a kind of a way of speaking about communicating ideas that are connected to the Judaism, to the transmission of Jewish ideas. And that that method of transmission isn’t arbitrary. That it matters whether you put something in Hebrew, that it matters that you put something in conversation with text from the past, that does something to the way that those ideas are received to the rhetoric around them, to the way that they’re influential.
And if we are at a moment now where the cost of taking an idea from non-Torah language and putting it into language is essentially zero. And if it’s not zero now, it’ll probably be zero soon. That does something that I think is threatening to the idea of the importance of that language because the language basically doesn’t matter. You can put anything in any language whenever you want.
So I’m, I’m curious if there are limits that you see kind of rushing towards us, if there’s things that you are particularly concerned about in the deployment of these technologies. And um, in ChatGPT specifically.
Sara: I do think that there’s a lot to worry about as well as a lot to be excited about.
Although I’ve also played the games of, you know, asking ChatGPT to write this and write that. It’s clear to me that the AI that we’re talking about at this point is really just a next step from what we had before. Like it doesn’t feel to me like a difference in kind so much as a difference in quantity, in some ways.
It’s not that different from typing a question into Google and getting an answer, which I’d say that you and I have been comfortable with for a significant portion of our lives. So that’s sort of one caveat to thinking about this. I agree with you about the value of language. I think that what ChatGPT challenges us to do, and really this is what Google challenged us to do, and this is what every new technology should challenge us to do, is to think about how do we educate for those things that we think are important?
I think even what you just said David Zvi, right, you’re sort of pushed to articulate that importance of language because you had this experience with ChatGPT, and if you were a classroom teacher, you could go back and have a conversation about that with your students, and that’s a really different choice. It doesn’t mean that schools are wrong in what I’ve seen, which is sort of this cracking down. I know that, one of my children’s schools sent out an email saying that henceforth, all writing assignments must be done exclusively in Google Docs with all version history visible.
So, okay, I get it. Also, they’re teenagers and teenagers sometimes make bad choices, and so great. Let’s do that if we feel that it’s necessary. I understand the frustration of a teacher who spends time grading an essay only to discover that it was written by ChatGPT, but at the same time, let’s have that thoughtful conversation that you just described, about why we think language is important and how we deploy it thoughtfully and effectively in the context of, let’s say, Torah study.
And let’s have students really think about that question. I think one thing to do might be to bring in the sermon about the bagels and have students critique it and really think about where are the pieces where we feel like this was great. This could jumpstart our learning. This could mean that we don’t have to memorize copious amounts of information, which I think is, you know, by the way, another really interesting conversation that was raised already by the advents of Google. Should it change what we have to memorize or not?
And then also a conversation about where did ChatGPT get it wrong? What do we notice about that? And what does that tell us about the role that we have to play in advancing learning?
David Zvi: That’s useful. The one piece of it that I struggle with is that there is an underlying assumption that this technology is here to stay, it is inevitable and irresistible. And so all we can do is basically, uh, accommodate. Even when it is incredibly disruptive to learning, even when it, you know, makes teachers’ jobs even harder than they already are.
And I guess my concern about that is the framework doesn’t allow for the possibility that there actually technological developments, developments in AI that do things that are actually bad, that are actually negative, that are not, cannot be explained away as morally neutral. And, you know, you have to use it right or use it wrong, but actually they’re doing something wrong. And I think there’s been a kind of intentional framing of ChatGPT and other AIs as being kind of societal projects, as a way to hide the fact that they are produced by specific corporations, by specific people who have profit motives, who have a certain vision for how the world’s supposed to develop.
But if you frame it as, no, this is just humanity doing this, this is just society doing this, then you can kind of avoid that. So I am curious about the classroom setting for another reason as well, which is that it is, and I think this is especially true for Jewish schools, it is not just a place where ideas are taught. It’s a place where morality is developed. And ideas about the world are set and ideas about technologies are set as well. So I am curious about whether you think that Jewish high schools, that Jewish schools in general should be places where some resistance to technologies like this where some resistance to technologies like this should be taught.
Sara: I definitely do, and that’s why I gave the example that I raised earlier, because again, I don’t wanna make teachers’ jobs harder. So if what you need to do is check the version history on the Google Docs, okay, I’m not in the classroom every day either. I’m not gonna argue with that.
But I really hope that that’s also being paired with thoughtful conversations. Why do we think this is bad? Like on some level, I think if I were a high school student, I have several who live with me, I would say, wait, like you just, you wanted me to produce the paper. For some reason it’s important that I produced the paper. The paper has been produced. Like what do you care how I produced it?
I’m a good user of ChatGPT, which is kind of like saying, I’m a good user of Google, which is kind of like saying maybe in a pre-Sefaria age, I’m a great user of the Bar Ilan “Proyect Hashut” website or DVD or whatever form you encountered it, right? Like why is searching, the ability to search and get the right answers, why is that not a valued skill?
I do believe that technologies are not necessarily neutral. I think that’s like a big maybe, you know, 1930s, 1940s, real belief, you know, technologies could be neutral and anything that was sort of scientific and efficient and, you know, we move in that direction, we’re moving away from sort of human emotion, and that’s not really true, of course. Humans are programming these things. Corporations are behind these things. We need to talk about that with our students.
And so to just say that technology, is bad, but inherently, I think in saying, don’t use it or we’re gonna catch you and out you as a cheater if you use it, we are judging it. We’re not skipping the judgment phase. I just hope that we’re having those thoughtful conversations. And as you know, I think that there’s a lot of Jewish language that we can use in framing what might be morally, ethically problematic about these new technologies.
You mentioned also in describing this sermon, the ideas that ChatGPT gets right and gets wrong. And one of the important choices that I believe that Sefaria has made in the development of our website is we have foregrounded primary texts. We’ve made that the backbone and the centerpiece of the site.
So again, going back to my idea that I think the ChatGPT exists on a continuum with these other technologies, I did teach in a high school classroom for a number of years. I did have the experience of students Googling answers to questions, and they would tell me, oh, I found this article, and this article says that Judaism says X. So it must be true.
And I think sure, you know, you could maybe make an argument that Judaism says almost anything, but, but my pedagogic philosophy is that I need you to show that to me somewhere in the primary text. I actually think that that’s a pedagogy that works for Jewish texts. I think it’s a pedagogy that works for all texts. I wouldn’t let you say something about Shakespeare without you being able to show me proof text also.
And so Sefaria in both putting the Jewish canon online and as much as possible in translation and making it available has made that possible. And also our insistence on linking between texts means that I hope that when students find things, even if they find an article, even if they find a secondary source, it links back to those primary sources, and so they can say where it really came from.
That’s a deeply Jewish way I believe, of learning texts. It’s a deeply Jewish way of talking about texts, and it’s something that maybe goes back to that question of language that you mentioned, right? That’s to me an authentic Jewish way of speaking to really cite your sources and know who said what. And that’s something that we can encourage the preservation of through the technologies that we develop and through the technologies that we choose to use.
David Zvi: I think that’s really powerful and um, there’s a couple things that came to mind when you said that. One is that in the moment when a student kind of googles something and kind of brings the world’s knowledge into the classroom, there’s something kind of similar going on when ChatGPT when you ask it some Jewish question and it returns an answer that cites a biblical verse or cites a piece of Talmud and that it is in a sense, kind of a distillation of what the internet knows about Torah.
And the internet knows a lot about Torah cause Jews have been on the internet a lot. So there’s something powerful about there. There’s something in that I, I think, could allow us to think about AI as a kind of kahal, as a kind of congregation, like a distillation of congregation and that you can kind of communicate with that as a whole. And that in itself exciting.
And the thing to think about is, um, just thinking about the development of skill sets, I know that people now are talking about how, you know, the new skill is not writing the essay. The new skill is developing the prompt, right? Like coming up with the right Google search terms that like, prompt development is its own new kind of skill that there’s something kind of similar going on. I, I know I certainly have this feeling, you know, when I was Yeshiva and kind of like constantly consulting Jastrow and other dictionaries, to try to find answers to words and thinking like, who was this person who came up with this dictionary. Like I can’t even imagine in my mind what it was like to be the kind of person who would write their own dictionary.
And the reason I can’t think of it is because like it was done. It’s done. And because it’s done, I don’t need to. The skills that I was trained on, that other people are trained on, are premised on the fact that that kind of knowledge is actually already easily available. And so it’s possible to move on to other things.
So there’s like something positive there. I’m still not entirely sold with the idea that like, there’s always gonna be a way to turn this positive that, you know, I think there’s deeper concerns here.
Sara: Can I say something about that?
David Zvi: Yeah, please.
Sara: I agree. I think that there’s a real negative, and I think the Jastrow example is so telling. I think it actually, it is kind of important, fun, interesting, informative to know who Jastrow was, maybe Jastrow was like not the best example cause of the dictionary, it’s like a little hard to make your personality come through in a dictionary.
But I think that there is this idea, which is a very Jewish idea, but certainly not only a Jewish idea, that personality, that worldview, that humanity, right, a person’s deep humanity informs their text. And I’ll give you a favorite example of mine that I think would kind of like, like, I don’t think ChatGPT would be able to do this.
And this is not the only place where it happens, but the Talmud at the beginning of Tractate Yevamot is dealing with a number of different cases about marriage, about levirate at marriage, and the Mishna lists a bunch of cases. The Gemara is trying to figure out why were the cases ordered in this way. So if you think about it, it’s very clear that the Talmud is relating to the Mishna as something that was written by a human being with preferences and emotions and real choices to be made.
And the Talmud eventually says, and I love this so much, that the example that comes first is one that was learned out through a process of Midrash, through a process of hermeneutics, of working through a text. And so the Tanna, the teacher of the mission loved that example more, and so put it first. And I think that’s an example of something that you could never say about a ChatGPT text.
ChatGPT does not love one piece of Torah more than another piece of Torah. I don’t understand enough about the algorithm. It was cited more times on the internet. It came up in more search results, right? But I think that that’s an example of something that we can talk about, we can teach about, is how personality deeply informs the Torah that we teach, the Torah that we craft. And that is a way of undermining the power of ChatGPT.
David Zvi: I’ve already seen actually um, some educators suggesting that the way to shift writing assignments is to personalize them. Not to say like, what does Shakespeare say about this? But like, what spoke to you? Which like, I think is both great, also has its own down sides. But I think that’s, that’s a, a promising avenue.
I’m also thinking about what you said earlier about the use of version histories as a way of authenticating that something written by a person was actually written by a person, that you could see that as a kind of, a technical check, but also understanding as a way of valuing process itself and imagining that as not just being a way of making sure you’re not lying, but actually something positive and something to be lauded and, and a way, and something to be looked at in and of itself.
Okay, last question. Artificial intelligence is a global phenomenon. It’s not just happening to Jews, it’s happening to everybody. ChatGPT effects all kinds of people in all kinds of ways. And there are certainly other religious communities that are dealing with these kinds of problems as well. What is your sense about how much of this is unique to Jewish education, in terms of the way that we are speaking about the impact of ChatGPT on the notions of holiness, on the way that sacred ideas are taught in religious context? How much of this feels unique to Judaism and how much is this just kind of implementing a just response to something that is inherently global?
Sara: I think that’s a really important question. I will say two things. One is I think that Jews have maybe an outsized emphasis on sacred texts. I think it’s not absent in other religions and some religions more so than others.
But we do have uniquely Jewish stances towards sacred text. We talked about scrolls, for example, as this technology that we’ve sort of oddly preserved. So I think we’re going to have to find a way to deal with all of this in a uniquely Jewish way, and I would say that the advent of digital texts was a very large, important first step in that.
Do you kiss your iPhone after you finish praying using your iPhone? That’s like the most extreme example that I give, but I think it’s something about this connection between the sacred text and sacred object that we haven’t solved for yet. I’m not sure that there’s a perfect solution that we should be looking for, but it’s something we have to keep thinking through. And so that’s one I think uniquely Jewish take or possibly uniquely Jewish take.
The other thing that I think is uniquely Jewish, is that we have a value that we place on Torah study, as we would say Torah l’shma. And so going back to that idea about the essay, if you submit an English essay or a history essay, it makes a big difference, how good is the essay? Did you get the answers right? Are they factually correct? Are they true to the story, et cetera?
When it comes to Torah study, on some level we would say, even if you came out with the best answers, if you only spent five minutes doing it, that’s less valuable than if you spent an hour doing it or two hours doing it. I think that one of the components of these technologies is they make accessing knowledge more efficient. They make producing new works more efficient, and we in Judaism are not necessarily going for efficiency.
And to tie that back to all these questions about pedagogy, I think that that’s a value that we can inculcate that we can really teach. I’m not sure that we do that when we tie Jewish studies so closely to grades. That’s like a larger conversation. But I do think that we have the opportunity to say no. It doesn’t matter if ChatGPT could give you all the answers in the world to all your Jewish questions. I want you to sit and study in the Beit Midrash with me for an hour.
David Zvi: Right. I think there’s a nice way to tie that to the idea of Torah l’shma, that Torah should be studied for its own sake, not towards some kind of end, but that the actual hour that you spend in the Beit Midrash has a value regardless of whether it is the most efficient use of your time.
And just on the, the kissing iPhones thing, the only reference I can think of outside of Judaism is I know in, in the late 19th century, early 20th century, there are Muslim discussions about whether recordings of the Quran on records, on, on photographs, have the same kind of holiness as an actual Quran should be treated somewhat separately. So there’s probably some interreligious work to be done there.
Sara: Yes, definitely.
David Zvi: All right, well, thank you for listening to the show. Special thanks to my guest, Sarah Wolkenfeld.
Identity Crisis is produced by me. It was edited by Gareth Hobbs at Silver Sound NYC. Our production manager is M. Louis Gordon. The show is produced with assistance from Miri Miller and Shalhevet Schwartz with music provided by Socalled. Maital Friedman is our vice president of communications and creatives.
Transcripts of our show are now available on our website, typically a week after an episode airs. To find them and to learn more about the Shalom Hartman Institute, visit us online at shalomhartman.org.
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, if you have comments about this episode, please write to us at [email protected]. You can also rate and review us on iTunes to help more people discover the show, subscribe to our show everywhere podcasts are available. See you next week. Thanks for listening.