8

How good does artificial intelligence need to be before we can turn to a robot for piskei halacha?

The accepted reason that even the most knowledgeable people need to ask legal advice from poskim is that the poskim possess special situational knowledge, experience, and sensitivity that most or all other people lack. Simulating [things like] these characteristics also happens to be the goal (or at least a milestone) of artificial intelligence research. When some AI reaches the necessary threshhold for humanness in these regards will it be reliable to issue individualized p'sak as well?

If not, will it be concomitantly forbidden to ask a question by email, over the phone, or by any other electronic means because of the concern that the answerer is pasul l'hora'a?

  • 3
    Do we have a sci-fi tag yet? :-) – Dave Feb 22 '11 at 5:09
  • Are you suggesting that we build our own Watson? – Tzvi Feb 22 '11 at 16:01
  • See also mi.yodeya.com/questions/6979 – msh210 Apr 24 '11 at 18:46
  • @Tzvi, No, Watson is the opposite of what this is all about. Watson works by executing very complex algorithms to come to a solution. I believe this question is asking about an AI that understands things in a human way. (Don't ask me if such a thing is possible. That's been a huge debate in the computer science, cognitive science, and philosophy worlds for decades). – Daniel May 8 '13 at 13:54
6

No amount of AI will be able to simulate Siyata Dishmaya. Heavenly assistance factors heavily into accurate psak. Just ask any posek.

A good posek also brings to bear a healthy intuition of how he would rule, even before he begins his analysis.

Think of a good physician with years of experience. As soon as she just looks at your child, she has a pretty good idea of what's wrong. This type of skill cannot be replicated by a machine.

Gary Kasporov also brings human intution into a decision when considering his next move on the chess board, which Deep Blue couldn't match. It only won with brute force (i.e. considering virtually every possible move before deciding). This is possible in chess where success/failure can be mathematically calculated. It doesn't work with pattern matching, which is what's needed in psak.

  • 1
    "It doesn't work with pattern matching, which is what's needed in psak." Have you seen the pattern recognition literature? They are working on that too. (Just one example: ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=4767034) How do you know that siyata diShmaya or at least intuition cannot be simulated given sufficient processing power and time? – WAF Feb 23 '11 at 1:46
  • Torah lives in the opinions of our Sages. See B"M 59b that the law follows the majority even when "wrong". This is certainly only true for human opinion, not digital opinion. – Barry Mar 8 '11 at 21:40
  • @Barry - So if the disqualification of a robot comes down to a binary exclusion of non-humans then the question shifts slightly away from artificial intelligence and toward artificial humanity. How close to human would a robot/cyborg need to be to be kasher l'hora'a (assuming, of course, the other prerequisites: mi.yodeya.com/questions/6979)? – WAF Apr 24 '11 at 19:14
  • 1
    Re Physician: Actually, there are computer programs that diagnose diseases, and they're pretty accurate. – Shmuel Jun 2 '14 at 6:49
  • Re Pattern Matching: Great progress has been made with image recognition, automated translation, and other pattern matching programs. – Shmuel Jun 2 '14 at 6:50
5

To support what Barry said:

  • There is an element of Divine assistance when the flesh-and-blood rabbi is called upon to rule (assuming he's qualified and doing his best).
  • We are easily decades (if not more) away from artificial intelligence vis-a-vis a lot of the skills that are needed for good psak. For instance, a young rabbi was training with Rabbi Moshe Feinstein; he saw two instances where a woman brought the same type of questionably-kosher chicken to Rabbi Feinstein, who prohibited the first and permitted the second. Rav Moshe explained that the first woman struck him as someone who wouldn't be stressed by throwing out one chicken, so he had her do so; but he could tell that the second woman desperately needed that chicken.
  • 1
    And why can't computers have divine assistance? – Shmuel Jun 2 '14 at 6:50
  • 1
    -1 This doesn't answer the question, which asks about a computer that has all the necessary skills. – Shmuel Jun 2 '14 at 6:52
-1

I think that the validity of a Psak is measured overall by its acceptance - the more it is accepted the more it is "valid" (or maybe even "true").

Sorry I forgot to answer your question:

The Turing test is just as good for Halachah as for any other field. Once people stop telling machines and Rabbis apart, Rabbi Google will become a legitimate part of our Jewish tradition.

In all fields of human knowledge, truthfulness, knowledgeability and efficiency of AI proved to supersede that of humans.

The only barrier from officially replacing Rabbis in Halachic judgments by an AI system is the Ruach Hakodesh stamp that we traditionally assign to them.

What will happen in [the field of] Jewish Halacha (just as in all other fields) is that at first, the Rabbis themselves will use the AI to assist them in their Psakim just as they start to use the computerized Otzar Hachochmah; second, they will acknowledge AI's intelligence and abilities and only sign their Haskomos to AI's Psakim; third, the Rabbis will become obsolete, just as doctors, lawyers, accountants, and judges.

I think that there's nothing in Judaism that truly contradicts that model besides our habits and tradition.

  • Did you mean to edit your other answer instead of posting a second one? – Monica Cellio Apr 2 at 15:38
  • @MonicaCellio I think they address two different non-overlapping aspects of the question. – Al Berko Apr 2 at 15:41
-2

An interesting and a bit complicated question. I think it is rooted in a deep misunderstanding of what a "Psak" is, to begin with. We're used to this word, but to apply it to AI we need to define it first ("see what-is-the-meaning-of-psak").

I answered it there - we need to differentiate a Psak (a verdict) and a Borerut (an arbitration).

It becomes clear that while the AI can be of great assistance in questions of "arbitration" - getting a sound and knowledgable advice from a know-it-all system based on personalized experience and Halachic fine-tuning, it cannot turn into a Posek as it lacks the intention of being a Psak (as I answered there).


Personal thoughts - I do feel that people will eventually cease seeking Psakim and be completely satisfied with personified bits of Halachic advice. As the Otzar Hachochmah and others will incorporate the AI in their systems they will arrive at the ability to provide customized guidance to all. It will become so easily reachable, professional, and indisputable that the idea of Ask you Rabbi will be forgotten. No Rabbi could compete with those systems and there will be a great shift in Judaism toward universality, unification, something we can start seeing today.

  • Are the two criteria based on your experience with Rav Shlezinger? Are you saying the "official verdict" in the Beis Din analogy is p'sak or is not p'sak? Re criterion #2, why can't a robot publish its conclusions? – WAF Apr 1 at 13:35
  • @WAF Thank you for reading. Yes, I could see his voice and body language and זיע ורטט when he was working on a Psak, as opposed to giving out "Halachic advices". – Al Berko Apr 1 at 13:41
  • @WAF This question is common to the supreme court judgements - when do they turn into a "law" and when it's just "local". The Gemmorah in Sanhedring makes that distinction - a Psak is obligating and a בוררות is not, similarly when a Rabbi gives you an advoce you still have a freedom of choice while if he gives a Psak you're "doomed" (if you truly follow him of course). – Al Berko Apr 1 at 13:45

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .