How the next generation of rabbis is using AI and ChatGPT – The Forward

Denise Blumenfeld’s AI learning tool doesn’t have all the answers — or at least it’s not so eager to give them. Instead, Blumenfeld, a second-year student at Orthodox women’s seminary Yeshivat Maharat, customized ChatGPT to respond to her with a question.

I recently saw Blumenfeld introduce a paragraph from the Talmud into the module, aptly named Socrates Havruta (his last name is Hebrew for study partner). His response tested his written comprehension: According to the first line of the text, what is the fundamental obligation regarding the lighting of candles? Blumenfeld typed a response, which Socrates confirmed before asking another.

She knew ChatGPT could just summarize the text, but would that really help her learn it? On the other hand, answering the questions might help someone figure things out on their own. And to be honest, she had a rule for herself: “I always try to read the authentic source first,” she told me.

Blumenfeld is part of the first generation of rabbinical students who are training with artificial intelligence tools like OpenAI’s ChatGPT and Google Gemini at their disposal — and hoping to prevent these tools from becoming a crutch. But the beit midrash, or study hall, is just one of many points of contact they have with an invention that could change not only what it means to be a student or a rabbi, but also what it means to be Jewish.

To get a sense of how AI is helping shape the next generation of rabbinate, I surveyed students from five U.S.-based rabbinical schools about how AI is used and thought about in their work. Their attitudes ranged from reserved enthusiasm to outright rejection. But their comments – and the boundaries they each set around their personal use – revealed the profound influence AI already has on their professional and religious outlook.

This was true even for students who did not use generative AI at all. Adrian Marcos, a student at the Ziegler School of Rabbinical Studies, listed the moral reasons he avoided it, including the exploitation of stolen data, its environmental impact and the digital culture crisis it accelerates. Still, Marcos admitted that the burden of explanation lies with AI’s detractors, not its supporters.

“A lot of people are very interested in AI, and as a rabbi, whether you end up in the pulpit or not, you have to converse with these people,” said Marcos, a second-year student at the Conservative seminary. “And as technology evolves, the conversations around it will evolve as well. »

Hack the sermon

ChatGPT can apparently draw on the entire digitized Jewish canon, translate it from Hebrew if necessary, and write new content on it. For students who pursue the rabbinate because of their passion for researching and sharing knowledge themselves, the question was whether a tool that lightened the load was truly useful.

Aiden Englander, a fourth-year student at Yeshiva University’s Rabbi Isaac Elhanan Theological Seminary, hires ChatGPT for intellectual on-call company. After finding an interesting commentary on the Torah portion of the week, he will ask ChatGPT for secular variations on the idea. It could spit out Nietzsche, he says, or a recent news story.

Knowing that ChatGPT can make such a connection opens up a different level of rhetorical ambition – serving as something of an academic force multiplier – but it precludes the possibility of sorting out its concept without it. This also raises the question of what makes a sermon “better” and whether literature you learned only through the ChatGPT summary is less suitable for a sermon than a volume you are still working on on the page.

Englander’s calculation was simple. “When you are able to quote a book that someone knows, they will remember it more,” he said.

Although he is an avid user, he harbors doubts about the reliability of AI. “She will completely constitute a Gemara,” he said, and in his opinion, her knowledge is particularly superficial when it comes to Jewish law. Yet what is ChatGPT can This is what makes the 24-year-old Englishman the most cautious. For the sake of creativity, he will ask him to point out some possible themes to explore in this week’s Torah portion to write a sermon. But it’s a muscle he’s aware he’s developing, so he tries to let go of it.

Yet a theological question about the use of AI persists behind utilitarian concerns. Is a ChatGPT generated d’var torah a bad idea because it risks spreading clichés or degrading a person’s writing ability – or because the very notion of computer recommendation defeats the purpose of the exercise, which is to apply human experience to Torah and vice versa? Major Jewish denominations have been as silent on religious questions surrounding AI as rabbinical schools have been on practical questions, leaving students to resolve both issues on their own.

“An LLM doesn’t have an autobiography – it’s not about having a faith experience,” said Dani Pattiz, a sophomore at Hebrew Union College. “He can come up with these brilliant syntheses of other people’s ideas. But ultimately he can’t truly glorify God, nor speak to people’s souls in an authentic way.”

A changing pulpit

As they navigated their own use of AI, rabbinical students wondered how it would reshape the lives of their future congregants and, in turn, their own work.

On a recent trip to Washington, Micah Glickman, a rabbinical student at Hebrew Union College, the Reform’s flagship seminary, visited a synagogue where a number of congregants had been laid off in the wake of DOGE’s federal job cuts. The impact of AI on employment, he realized, could be exponentially greater. If that were the case, it wouldn’t just mean more people would face financial and emotional vulnerability. It was also about the fact that a universal source of human flourishing might have an expiration date.

“It seems like the promise of this technology is to do basically anything a person can do, and do it better than that person can,” Glickman said in an interview. “And I wonder how that will affect a congregation of people who perhaps derive meaning and purpose from their accomplishments in life.”

It would be up to rabbis and other religious leaders, Glickman said, to shepherd their communities through this change. “We are on the verge of an imminent spiritual crisis,” he said. He was helping organize a symposium at HUC to examine these and other questions related to AI – seeking a spiritual solution, he said, to a spiritual problem.

Meanwhile, he was already encountering bar mitzvah speeches written by ChatGPT from the students he teaches. This put him in the position of his HUC teachers: should he discourage children from using it, or – assuming it is inevitable – try to steer them towards responsible use of AI?

This was a recurring theme in the interviews: The future rabbis I spoke with were more worried about how the generation that came after them would learn than about their own trajectory. After all, today’s students have largely succeeded in their studies before rabbinical school without ChatGPT; they were wary of atrophy only because they knew they had muscles to begin with.

Even Blumenfeld, the Maharat student, wasn’t sure whether to recommend her Socrates robot to younger students. “Because I had experience learning and teaching before AI, I know what the outcome I’m looking for is and I know how to ask for the right thing,” she said, while children at this age had not yet developed these skills. “As teachers,” she added, “we must learn how to teach. »

They must also learn to learn, and Why learn. The YU Englishman recalled a thought experiment shared decades ago by former university president Rabbi Norman Lamm: If you could implant a microchip in your brain that would give you complete knowledge of the Torah, would you ever need to learn? Lamm’s view was that learning actually took precedence over knowledge – that the labor of Torah study was not just a means to an end, but a form of worship in itself.

“It’s perhaps a bit more of a mystical notion, but from the point of view of accessing Hashem, it’s only done by learning the text and wrestling with it, without being told what the text says by a third party,” Englander said. The reason he could usually detect when ChatGPT was hallucinating the Talmud, he added, was because he had devoted hours to studying it himself.