With the rise of AI, will we have robot rabbis?

Question: Artificial intelligence is advancing so quickly that before we know it, we might have “artificial rabbis”. Would it be permissible to refer halachic questions to one?

Rabbi Brawer: The following story is told of the famed Rabbi Shmuel Salant, who served as the Ashkenazi Chief Rabbi of Jerusalem for almost seventy years. A woman once approached him with a sheilah, a halachic question, concerning the status of a stew she had cooked.

It transpired that the meat she had put into the stew was not kosher, what should she do? The rabbi advised the woman that so long as she removed the offending piece of meat, the remaining stew was fit to eat.

Some weeks later, the woman related her experience to a friend. The friend was incredulous, for she had asked the rabbi the very same question previously and the rabbi’s answer was entirely different; he instructed her to throw out the stew, and clean and purge the pot.

When the rabbi was confronted with these seemingly contradictory rulings, he replied that situational circumstances guided his halachic determination. He went on to elaborate that the woman who received a lenient ruling was poor. Furthermore, she had a large family. He was able to deduce from this she could ill afford meat and so the offending piece would have been relatively small, and the woman would have bulked up the stew with cheap potatoes. Given the size of her pot (she cooked for a large family) and the ratio of meat per potatoes, the flavour of the non-kosher meat would have been cancelled out.

The other woman, however, was affluent. Furthermore, she had no children. The rabbi deduced from this her pots were smaller, and the ratio of non-kosher meat per potatoes was much larger, hence his more stringent ruling.

The point of this story is that Rabbi Salant was an effective rabbi, not only because of his halachic erudition, but also because he really knew and understood the human being behind the questioner. A halachic decisor does not just compute facts, but rather tries to build a comprehensive picture of all related circumstances. Being book smart alone does not a good rabbi make.

A similar approach is gaining traction in the medical field, captured by the adage, “Treat the patient, not the disease”. Physicians are being encouraged to look beyond the specific ailment, to gain a fuller picture of the whole patient.

Computers are binary. Human beings are complex and multifaceted. Halachah is an all-encompassing entity. It responds to and shapes the human experience. As such, only a human being can interpret and apply it effectively.

Rabbi Brawer is Neubauer chief executive of Hillel, Tufts University

Rabbi Romain: In some respects, the robo-rabbi is already here (also operating under the name of Rabbi Google), as there are many electronic ways of finding not just simple information, but delving into a vast array of responsa from highly authoritative sources.

For those used to shuffling from one book to another, or having to go to a specialist library for a rarely quoted opinion or an out-of-print text, the ability to consult widely and at speed is both astonishing and welcome.

However, as with artificial intelligence in general, there are limitations. Yes, you could get factual information from an artificial rabbi, which would also be able to deal with subsequent questions or any qualifications arising. However, would it be able to deal with highly personal nuances?

If someone asked about the afterlife, for instance, could an artificial rabbi tell if they wanted a survey of Jewish views on the subject, or if they were hinting at their own fears of the unknown, or needed to talk about their angst about what is happening to them in this life?

There is also the question of what angle the artificial rabbi might take when delivering a judgment? Would it be totally neutral and cover all options, or guide someone in a particular direction, be it Orthodox or Progressive?

Much would depend on how it was programmed and who designed it. Might this lead to different types of AI on the market: from OrthoBot to ProgRob?

Of course, for Jews, the concept of artificial intelligence is not new, as we already have a Golem mentioned as far back as the fifth century in the Talmud (Sanhedrin 65b), and which then became the subject of legendary tales in the late Middle Ages.

It may be relevant for us that, despite its amazing powers, all such stories conclude with the fact that eventually the Golem runs amok and has to be destroyed lest it endangers human life.

It could also be said that humans are God’s AI: we were formed in the image of God and are carrying on God’s creative abilities. But we too went out of control, disobeying God, eating from the tree of knowledge and left Eden with mixed results for God’s world, which are both wonderful and horrendous.

It seems that AI can carry many advantages, but also huge dangers, and we have to be sure we can regulate and control what emerges.