Parashat Yitro: Of Rabbis and Robots, and the Artificial Posek

The dawn of the Jewish People’s relationship with the study and instruction of Torah is relived this week with the reading of Parashat Yitro. We do so, of course, in a radically different world, one that is as we speak being transformed by the rapid advancement of Artificial Intelligence technology, prompting complex questions that span technical, practical, sociological, and deeply spiritual realms. This is particularly salient for Judaism, with its age-old tradition of venerating scholarship, a goal in and of itself as well as the prerequisite to determining the guidelines of Jewish practice, the principles and rules that instruct everything from the most mundane aspects of daily living to the loftiest issues of ethical complexity. Will the “People of the book” become the people of the computer? Will rabbis be replaced by robots?

Currently, there are programs capable of answering inquiries about Jewish law, mimicking the Responsa literature that has guided Jewish communities for centuries. Scholars are evaluating this phenomenon: What part can a machine play in this process? Can it supplant the human role entirely? Could it serve as a supplement or an enhancement? Or should it be excluded completely?
Classical Hebrew has many terms for intelligence, and Modern Hebrew has chosen the word “binah” for translating Artificial intelligence. This may reflect what aspect of intelligence is being “created”, as this term is possibly best rendered as “understanding”, more specifically, “understanding one thing through another.”

The Talmud also indicates that binah is achieved through wrestling with error. If so, can a machine ever attain this quality? Current iterations of generative AI make many mistakes, and even learn from these mistakes, thus refining their answers. Still, is this what is meant by understanding emerging from error? Or do we speak of something more fundamentally human, the depth of wisdom that is only obtained through adjusting one’s perspective due to the encounters with the peaks and valleys of real life?

The quest for Artificial Intelligence has always been as philosophical as it is scientific. Pioneers like Norbert Wiener as early as 1948 anticipated risks, warning that information technology might elude human control and make decisions based on utilitarian calculations rather than human values. Early AI development followed two tracks: those seeking to recreate human intelligence and those aiming to augment human capabilities.

AI’s breakthrough came with neural networks, which process through structural relationships rather than predetermined symbolic rules. Current generative AI models like GPT use large language models (LLMs) to encode statistical relationships. “GPT” refers to pretrained “transformers” that can assess the significance of the data they have been trained upon, and generate new content. But whether this constitutes genuine intelligence remains debatable. Some have provocatively defined intelligence as “whatever computers can’t do yet.”
The question is profoundly significant in Judaism, as some thinkers, notably Maimonides, see intelligence as defining the Biblical notion that Man is created in God’s image. There are additional theories, too, including creativity, the ability to handle complexity and contradiction, to experience emotions, and to care about others and motivate altruism. Aspects of these possibilities appear in secular theories as well of defining intelligence, or at least the intelligence that truly matters.

However, intelligence is insufficient to qualify as an authority on Jewish law. The Talmud makes it clear, in a widely cited passage, that God has chosen to involve humanity in the halakhic process, and in fact to grant the finalization of that process to them, as it is no longer “in Heaven”. Presumably, this role should not be ceded to a computer; it thus behooves us to identify what part of the halakhic process necessarily must be preserved for human involvement.

The Talmud relates that when Moses went to Mount Sinai to receive the Torah, he met opposition from the Angels, who argued that the Torah should not be placed in the hands of mere mortals. Moses argues to defend his people, and ultimately prevails, asserting that it is only human beings who have the physical needs that the commandments of the Torah speak to, eating, drinking and like.

There is, as some note, something difficult about the story. Why did the Angels care about the Jews receiving the Torah? In what way would that impact their own possession of its treasures? Why not allow it to be the case that both the angels and the mortal Jews should be permitted the beauty of Torah study?

One suggestion to resolve this issue, building on the words of the medieval scholar Rabbenu Nissim, is to note that the handover of the Torah to the Jewish people at this stage reflected much more than the license to study its content: it represented the partnership that God was making with his people, to determine the direction of Jewish law, a trust that can only be given to the human beings who live the lives the Torah talks to.

This is among the objections raised against the possibility of machines rivaling human thought. As Daniel L Everett has put it, “the more we learn about cognition, the stronger becomes the case for understanding human thinking as … the interaction of the human body, human emotions, culture, and the specialized capacities of the entire brain…. Computers may be able to solve a lot of problems. But they cannot love…”

Talmudic texts also debate whether erudition or analytical ability is more important in rabbinical leadership. If either could be identified as superior, it would address the role technology could play in mimicking that attribute. The conclusion apparently favors the knowledgeable over the incisive; however, in the nineteenth century, some authorities asserted that the Talmud’s conclusion was no longer in effect, now that the printing press made religious texts widely available for immediate consultation. Others denied that such early “outsourcing” of erudition could diminish the importance of a deeply learned human being. Either way, the existence of digitally available information may reshape the role.

This question combines with the fundamental issue of how Jewish law is decided: while erudition is crucial as a starting point, the process of deciding is identified by one dominant cognitive act: “comparing one thing to another.”

This parallels what Douglas Hofstadter argues is the dominant definition of intelligence, after surveying countless other suggestions: the power of analogy, “the ability to pinpoint the gist of a situation in a flash… to get quickly at what matters and ignore the rest…the art of, when one is facing a new situation, swiftly and surely homing in on an insightful precedent (or family of precedents) stored in the recesses of one’s memory.” At first glance, this would seem to describe the process by which GPT generates output. However, some computer scientists argue that AI fails to make meaningful cross-domain analogies in this sense.

Debate rages over how AI mimics, or potentially undermines, human intelligence. Noam Chomsky claims that machine learning degrades our understanding of thought, reducing it to mere statistical pattern matching. In response, Salman Kahn suggests that human creativity often stems such matching, arguing that collaboration, even with AI, can enhance creativity.

Similarly, Yuval Noah Harari has written that he anticipates computers overtaking humans in religion, just as they have in chess. David P. Goldman, in the journal Hakira, strongly disagreed, writing that chess, ”is a game with fixed rules, while [Jewish law] is infinitely open-ended, always modified by … human creativity. … Equating men to machines is inimical to Judaism, whose four millennia of history and practice stand as a reproach to sophomoric speculation of this kind.”

There is a middle ground: humanity has unique creative power, but it is rooted in the ability to compare one fact to another. Increased exposure to data, and also to possible distinctions, is the training ground for the creativity cherished in the deciding of Jewish law.

Still, numerous other issues abound: Jewish law emphasizes the personal stature of a particular authority; can a machine qualify, or properly assess the relative rankings of the authors who opinions it trains upon? This is a general issue; Jeremy Kahn, in Understanding AI, argues that here, AI is a step down from the internet, in which “reading multiple links introduces a modicum of friction which … provides space for critical though [calling attention to] where information is coming from… generative AI removes this friction, and in providing a confident, summarized answer, makes the abandonment of critical thinking all too easy.” He argues that paradoxically, easier access to information has engendered a narrowing of science and scholarship.”, while older research “facilitated broader comparisons.”
Chess master Gary Kasparov, discussing his defeat by a computer in his book Deep Thinking, writes “Using our machines to acquire, and retain more knowledge cannot be a bad thing on its own. …the problem comes when the database and the engine go from coach to oracle.”

Further, AI is vulnerable to deliberate, or even subconscious, manipulation. Steven Brill, in The Death of Truth, discusses how AI can produce a negative or positive review on demand, calling into question the significance of either. Similarly, developers of Jewish law AI technology acknowledge that it can provide either lenient or stringent conclusions upon request. This highlights the necessity of human involvement; an actual person will need to decide what decision is more compelling, or appropriate for the circumstances.

Most concerning are “hallucinations”, simply false responses, often disturbingly plausible, that can emerge from generative AI, seemingly a fatal flaw. How much this can be eliminated – as current efforts are addressing – remains to be seen.
This all goes to the narrow question of AI addressing the abstract legal issue. Additional reasons why, in a practical sense rulings of Jewish law could not be left to AI, chief among them the importance of the human elements involved in addressing specific situations and questioners.

From a character perspective, the Talmud emphasizes humility as crucial in prevailing in disputes. How could a computer incorporate this? Further, it is believed that Divine guidance is provided to those with personal piety who have taken on the responsibility of advising their followers. Can this be replicated through computers?

Surprisingly, some feel that it can, that personal merit can be transferred through the human input leading to the machine’s answer, or perhaps through the one consulting it. This raises questions regarding the nature of the partnership between humans and AI, a topic addressed in an essay by Samuel Arbesman with the felicitous title. “Nachas From Our Machines”.

The biggest concern is the human empathy required of authorities whose rulings are most valued. Here, it should be no surprise that computers fall short. A report from New Scientist noted, regarding a study done on utilizing ChatGBT in the legal system, that “the AIs had none of the understanding or empathy we might expect from human decision makers”.

Philosopher Thomas Metzinger has written, “We’re smart because we hurt, because we can regret, and because of our constant striving to find some viable form of self-deception or symbolic immortality.” Further, Maria Popov argues “Morality is predicated on consciousness and on having a self-conscious inner life rich enough to contemplate the question of what is ideal.”

Ultimately, Jewish law has always been about the Divine conception of human potential. To borrow from Lincoln, it starts in Heaven, but must still be “of the people, by the people, and for the people.” The dazzling technology of AI can make enormous contributions to the process, and the sky would seem to be the limit. But Jewish law is not decided in the sky. The image of God remains a uniquely human gift—complex, compassionate, and continually growing.