In November 2022, OpenAI – a non-profit organization – released a free preview of ChatGPT. This triggered widespread interest in generative AI and leading to the ongoing AI boom. By March 2023, OpenAI released a new model – GPT-4 – and the update race began in earnest.
One of the first Hasidic groups to response to the rapid growth in AI interest was the Beth Din Zedek D’Chasidai Square – the rabbinic body of Skver Hasidim responsible for Jewish Law.
Skver Hasidism comes from the Chernobyl Hasidic dynasty. Chernobyl traces its lineage back to the earliest days of the Hasidic movement, and this dynasty begat many Hasidic courts. Skver is originally from Skvyra in central Ukraine. Today the Hasidic group is based in New York. The largest branch of Skver Hasidism is headed by Rabbi David Twersky (born 1940) and is based in New Square in Rockland County, New York.
On April 27, 2023, in the lead up to the festival of Shavuot, the Skver Beth Din issued a public statement that included an urgent warning against was it called “the new service ‘Open AI’ and the like, which could be used via computer, or through text (perhaps meaning phone text – LC), or even through a simple telephone.”
The notice alerted readers that at this moment in time, it was possible that not everyone was aware of the “great danger” that AI poses. The Skver Beth Din explained that it had investigated the matter and established that “without a doubt it is a terrible threat and great danger for all – old and young – may God have mercy.”
The warning does not explicitly say why OpenAI was deemed so dangerous. At this early stage its capabilities and potential were not widely known, so at first blush it is not immediately apparent why the Skver Beth Din was so alarmed.
It seems that the Skver Beth Din assumed that AI was something akin the internet, rather than a radically different tool. Indeed, the Beth Din called the service “an open internet without any filter,” though those words appeared in inverted commas. Given that AI was perceived as filter-less internet, the Skver Beth Din charged that it was open to all kinds of depravities and heresies.
“It is therefore simple,” continued the notice, “that AI is included in ‘the grave prohibition’ against internet without a filter that has been asserted by the great rabbis of the generation.”
In large bold lettering – for those who would not pause to read the entire text – the notice declared: “The use of AI is utterly forbidden, in any way and form whatsoever, even via a telephone.”
The statement concluded with a prayer that God should “guard us with a watchful eye, to purify us and to sanctify us with lofty holiness, and we should merit to raise our descendants as our souls desire, and receive the holy Torah with all the good endowments.” Seventeen signatures were affixed to the bottom of the notice.
One journalist asked ChatGPT (not the OpenAI organization) what it thought of being banned by the Skver Beth Din. ChatGPT was unfazed by the diatribe, acknowledging that it was powerless to take action against public attacks, and contextualizing the document: “The ban stems from concerns about the negative impact of technology on the lifestyle of the ultra-Orthodox. It is important to note that each community has the right to set its own guidelines and rules, and it is the decision of the individuals within the community whether to act on these guidelines.” ChatGPT also reminded users of its role: “I exist solely as a tool to provide information and assistance to users, and I have no ability to make decisions or take actions that would cause controversy within a particular community or group.”
Nowadays, such a ChatGPT response is unremarkable. But back in 2023, Ynet – the online outlet of the Yedioth Ahronoth newspaper – considered ChatGPT’s response a newsworthy moment.
****
The Hasidic response to the ban was swift: Within a few days, an enterprising Chabad Hasid, Moishy Goldstein, created Kosher.Chat – a chatbot based on OpenAI that was trained to be sensitive to the concerns of the Skver Beth Din.
Kosher.Chat was essentially a filtered ChatGPT. Officially, it would only answer non-Torah questions. It would not relate to zmanim – times for prayer. Nor would it authoritatively state what was the Torah’s official stance on issues. Kosher.Chat also declared that it would not authoritatively answer halakhic questions. For halakhic questions, the chatbot directed users to various online Chabad outlets. Notwithstanding, when asked halakhic questions, Kosher.Chat would provide an answer.
Since 2023, other filtered Large Language Models (LLMs) have been released, including Kosher AI, KosherGPT, and RavGPT. It is not entirely clear whether there is rabbinic oversight and by what standard these LLMs are “kosher.” The underlying theme of these models is that unfiltered LLMs are treif.
****
There is an alternate way of thinking about Large Language Models that is ground in Hasidic thought. Rather than using the kosher/treif dichotomy, perhaps it would be more appropriate to turn to the idea of Kelipat Noga – a Kabbalistic idea given prominence in Hasidic writings (see, for example, Tanya, Ch. 6-7; Likkutei Mohoran 1:19 and 82).
According to Jewish mystical tradition, there are kelipot, shells or husks, that represent darkness, impurity, and evil. The kelipot serve as an obstacle that makes it difficult to perceive the divine presence in our physical world. Things that are forbidden according to Jewish law – such as pig, or LLMs according to the Skver Beth Din – draw their lifeforce from kelipot and therefore should be completely avoided.
One kelipa – Kelipat Noga, a glowing husk – is substantially different to other kelipot. This shell has mystically-neutral valence, that can be used for holy purposes or – Heaven forfend! – deployed for evil. Eating, working, speaking – are all physical actions that engage Kelipat Noga. In an abstract sense, these acts have neutral spiritual significance. Their spiritual destiny is undetermined, until humans decide whether to sully these husks with impurity, or whether the husks will be elevated so that they glow with sanctity.
It would seem that LLMs could be a classic case of Kelipat Noga – a mere husk that is neither good nor bad in an apriori sense. AI could become a destructive element in the world, or it could be a positive, constructive force that contributes to humanity. The spiritual destiny of Artificial Intelligence is entrusted in human hands.
The writer is a senior faculty member at the Pardes Institute of Jewish Studies, a senior lecturer at Bar Ilan University’s Faculty of Law, and a rabbi in Zur Hadassa.
