Over the past year, I have watched as Jewish educators have become increasingly interested in AI. Developers and amateurs alike are creating apps trained on Jewish texts, introducing us to avatar rabbis and personalizing Jewish learning in entirely new ways. Just last week, Sefaria released a new translation of a biblical commentary, 98% of which was generated by Anthropic’s large language model, Claude.
It is still unclear what these changes will mean for the future of Jewish education. We don’t yet know where AI can add real value and further the goals of Jewish learning. This is a time of experimentation, and I’m here for it.
However, there is one area that AI must never go: the Shabbat morning sermon.
If it’s just for research? Go ahead. For the actual writing? Stay away.
Keep My Jewish Learning free. In this season of freedom, help us keep Jewish connection and meaning free for everyone. Your support ensures that anyone seeking Jewish wisdom can find it here, without needing to cross the red sea (or a paywall).
Before I explain why I feel so strongly about this, I need to say: I get it. Being a rabbi is a tough job, and inspiration does not strike with regularity each week. Time spent on sermon writing is time away from congregants, and it makes sense that rabbis would want to prioritize the latter.
It’s all the more tempting because the downside isn’t so clear. AI sermons are actually quite good; this was true even before ChatGPT, and they’ve only gotten better. In the past four decades, Jews have filled the internet with an immense amount of Jewish content that has now been hoovered up by Gemini, ChatGPT, Claude and even Grok. Why not tap into this wisdom and save a little time?
Well, here are three reasons.
- Religious AI policies need teeth to be meaningful
I’ve spent a lot of time this past year observing as religious denominations of all kinds try to set internal policies on AI use. These processes, while sometimes producing elegant language, often balk at the last, crucial step: What do we actually want people to do differently?
Since the AI transition feels inevitable, telling people that a Christian or Buddhist or Jew ought to use AI in some particular way can feel futile. Worse: Religious leaders may already worry about being out of touch with the public, and saying no to a technology like AI — even in a small way — can feel like the perfect strategy for sustaining that reputation.
But if your AI principles don’t dictate action, what’s the point of having them? It’s all well and good to develop a religious approach to AI, but if that approach never asks anyone to do something against their economic interest, in what sense is it religious? Faith-based AI policies should give people the confidence to act out their values when doing so carries a cost. If they don’t, there’s really no reason for them to exist.
How can religious leaders exemplify this approach to principled AI use? By modeling it themselves. A no-AI policy for sermon-writing is easy to understand, easy to implement (though perhaps not to enforce), involves relatively minor sacrifices and is highly visible. If you want to send the message that principled AI use is possible, this is the place to do it — and if you can’t set an AI policy here, that doesn’t bode well for your ability to set effective tech policy elsewhere.
But do people really want AI out of their sermons? The answer, probably, is yes.
- People don’t want AI in synagogue life
Despite prognostications about religion driven increasingly by AI and the much-hyped rise of Jesus bots, a large majority of Americans do not think AI should be advising people about “faith in God.” Religious leaders who worry that rejecting AI will make them seem out of touch might have it exactly backwards. All of us are awash in a sea of AI slop. When we go to services, we mostly want to connect to God and other people. We want something real.
We also want to connect in person. Even after the pandemic, American church-goers still prefer to attend services in an actual building. Since AI technology is still largely restricted to virtual settings, physical space also serves as a natural setting for human-to-human connection — and this is especially true on Shabbat, when many Jews restrict their use of electronics. On a day and in a place that is already naturally devoid of artificial intelligence, a policy of human-written sermons can reinforce the image of the synagogue as a place for people alone.
- This is a very Jewish investment in the future
The decision to accept or reject a new technology often feels small in the moment; its true significance often takes years to fully emerge. We live in a tiny historical window in which there’s a large overlap between AI and human writing abilities because few people (recent high school and college graduates being the exception) were trained in an environment where AI was available. This gives us the illusion that the switch to AI-written sermons is both trivial and reversible, but it is neither. Skill redundancy can only last so long; if AI sermon writing becomes the norm, the human skill of sermon-writing will inevitably wither.
Judaism is big enough to hold onto old technologies, and doing so can create new opportunities for meaning-making. Consider that the Torah scroll was once just a very large book; it became notable only after the widespread adoption of the codex. As manuscripts gave way to print, the scroll’s distinctiveness grew. Today, the Torah scroll is one of Judaism’s most recognizable symbols. Yes, the fact that they are written by trained scribes means that they are quite costly, but you’d be hard pressed to find a Jewish leader who thinks the expense is cause for their elimination. Scrolls are antiquated and inefficient, just the way we like them.
We don’t yet know the long-term effects of AI on how the public writes or reads, but it is easy to imagine a near-future world where the allure of quickly produced, good-enough prose mostly outweighs the time-intensive exercise of stringing together words manually. People may grow impatient with the stop-and-start nature of the creative process and choose to forgo deep connection to their writing for the sake of greater output. In doing so, fewer and fewer people will choose to engage with writing as an art form.
But what if the sermon preserved that art form? It is no small thing to find something inspiring to tell a community every week. The frustration of finding the right words, of discovering the most interesting source material, of deciding what it is you want to say, of not getting it quite right, of literally writing or typing out all the words — I suspect that we will come to cherish these inefficiencies in a world that no longer has much use for them. Just as we hold Torah scrolls in high regard despite the fact that almost every Jewish book is available instantly, we may come to hold dear those few minutes each week when we hear a message crafted by and for humans, even if AI could have said something possibly better and definitely faster.
The human sermon can become a new part of Jewish identity. All we need to do is put our foot down and just say no.
