YeshivaThe torah world Gateway
Beit MidrashAsk the RabbiCalendarWikiDedications
Eve of Thursday 23 Menachem Av 5786

Dedicate the study in the website

Dedicate

YeshivaThe torah world Gateway
 Beit Midrash

Beit Midrash
  • Sections
  • Chemdat Yamim
  • Bemare Habazak - Rabbis Questions
קטגוריה משנית
undefined
0
Shorts AreaFemale LessonParashat HashavuaSeries'LibraryPiskei Din (Hebrew)Test Yourself
Question: To what extent is it responsible and wise to use AI models for guidance or information on questions of practical Halacha? Answer: This discussion is for those for whom AI is already part of their practical, intellectual, and/or professional lifestyle. What follows is in line with the broad consensus amongst rabbanim in our communities. Hopefully the presentation will clarify somewhat and highlight distinctions and nuances. A futuristic possibility – Someday, machines might surpass humans in sheer memory, analysis, and application of Halacha (if they do, it will happen later than for most other intellectual endeavors). We must distinguish between halachic information and p’sak halacha. When a halacha is clearly established and a non-expert does not remember facts or know how to “connect clear dots,” the future AI will be trusted. For time immemorial, we do not ask our posek such questions, but ask a more knowledgeable acquaintance. However, when a new question or a unique dilemma including conflicting factors or personal sensitives arises, this requires a p’sak halacha. The process that the Torah describes for deciding such matters (Devarim 17:8-11) involves human talmidei chachamim, which outweighs even a divine omen (Bava Metzia 59b). There is also something to be said for siyata d’shmaya of a yarei shamayim toiling over an answer. Current AI capabilities – As many know, while AI does certain things in the realm of Torah impressively, its level of reliable scholarship pales in comparison to even a basic-level talmid chacham. AI summarizes straightforward, basic halachic text well. (One AI model retrieved for me my positions on a variety of topics quickly and impressively.) Some of its most impressive feats are plagiarism – summarizing articles by talmidei chachamim without citing them. It can follow Talmudic and halachic logic and notice distinctions. One can get a broad, brilliant answer on a halachic question. However, it is totally unreliable. It can “quote” non-existent sources. It can “miss” a part of the question and apply an answer in its “grasp” about a similar case; i.e., the answer is wrong for the actual case. It can give an answer based on one source it understood correctly, but fail to realize that most opinions disagree. I compare it to having an absolutely brilliant chavruta who is new to the world of Torah. When it gets something right, it can be breathtaking. However, when that lack of proper background causes it to err, it can say things that sound great but are totally off. My impression is that on “cookie-cutter” clear matters, where there is little complexity and no machloket, it usually gets things right. The chances are higher when one knows how to ask AI in the right way. Moments ago, I asked three AI models the berachot on 7 foods (quite straightforward). They all were accurate about the foods that have one correct answer. The answers for foods upon which there are machlokot differed from model to model, as did the level of accuracy. Here is my practical summary. A talmid chacham who lacks encyclopedic knowledge, but knows how to check everything before believing anything, can, if he desires, use AI to get to opinions and approaches. He can be aided faster and sometimes more efficiently than by using a Google search to find and read articles or reading through classical sources, which he might find “the old way” or with the help of Bar Ilan Responsa, Otzar Hachochma, etc. It serves more like a halachic anthology sefer with an “automatic index” when you don’t look up the sources cited in footnotes. The researcher must avoid the danger of falling into the horrible habit of believing too much, which already exists regarding some anthology sefarim; it certainly exists for AI. To quote one AI model’s disclaimer: “XXX is AI and can make mistakes. Please double-check cited sources.” For those who are not talmidei chachamim, using AI to get Torah information is usually highly irresponsible. (This is true for non-experts regarding other intellectual or professional fields.)
Bemare Habazak - Rabbis Questions (654)
Rabbi Daniel Mann
656 - Ask the Rabbi: Kittel at the Seder
657 - Ask the Rabbi: Reusing a Grave
658 - Ask the Rabbi: Sefirat Ha’omer of Someone Who Does Not Understand
Load More
undefined

Rabbi Daniel Mann

Bemare Habazak - Rabbis Questions
Bemare Habazak - Rabbis Questions
Popular Lessons
Popular Lessons
Recent Lessons
Recent Lessons

Rabbis in Midrash

Lessons
fast navigation
עוד Lessons
Personal area

Join now to watch your latest lessons and questions and continue watching series' of lessons right from the time you left them.

את המידע הדפסתי באמצעות אתר yeshiva.org.il