[Avodah] AI?

[Micha Berger]

On Wed, Oct 30, 2024 at 05:30:40AM +0200, Joel Rich via Avodah wrote:
> My comment on an AI panel:
> Most of the professionals that I've heard speak on this topic are pushing
> the joint model where the AI assists the professional with input, but the
> professional makes the final decisions. What's always interested in me in
> this model is at what point will the professionals lose some of their
> native abilities...

I don't think  this is a real threat. The use case you are describing
is forced by how often AI is wrong. "Hallicinates" as they fancifully
put it. So you need an expert to decide what's usable and what isn't.

Picture if you owned a calculator that was wrong 25% of the time. And
because the error could be in any digit, knowing that the answer is
plausible isn't enough to be able to use it. You need to have full
expertise in identifying correct answers in order to know when to
accept a result and when not.

The AI assitant model is only worthwhile in domains where checking an
answer takes a lot less work than producing one. Like the difference
between proofreading and tweaking a letter vs writing your ideas down
in letter form.

So I think the primary skill lost is looking for sources. Which was
already changing radically because of online librariews and search
engines.


R Dr Moshe Koppel is in the middle of a series of blog posts which use
archetypal members of various O communities to frame a dialog about the
nature of halakhah, and what that nature ought to be.

Interestingly, AI came up in today's installment:
    Can AI Replace Your Rebbe
    https://substack.com/home/post/p-150935589

Teaser:

    Yoily clears his throat....

The stereotypically Satmar name is relevant. Continuing a bit down,
I just through that in to explain the following "he":

    He pauses, taking in the dubious looks. Someone shifts in their
    chair, and Yoily gives a shrug. "Our host mentioned all the changes
    we need to adjust to. But he didn't mention the biggest change
    of all. AI, I'm talking about AI. Artificial intelligence, if you
    need it spelled out. Why? Because it's objective, it doesn't play
    favorites. It doesn't care if you're a chossid, Sephardi, Litvak,
    or some mix of everything. You ask a question, and it gives you an
    answer. A clear answer, no arguments, no haggling."

    The host raises an eyebrow, clearly intrigued. "Go on, Yoily. We're
    listening."

    "You see, the problem with Yiddishkeit today," Yoily continues, "is
    that everything's subjective, everyone's got their own rebbe, their
    own community, their own way of doing things. It's chaos, and that's
    fine when you're a small group, but we're not small anymore. You
    can't run a country like that. What we need is one standard. And let
    me tell you, AI can give it to us. You take all the halacha, all the
    psakim, the minhagim, everything, and you feed it into a machine. The
    machine will give you the answer. Not an answer, the answer."

    He pauses, his eyes sweeping the room. "And before you start with,
    'Oh, but halacha is so complicated, it's got nuance,' believe me,
    I know. But that's why AI is perfect. You feed it the nuances, the
    exceptions, the whole tangled mess, and it finds a way through. It
    finds consistency where humans just find arguments."

    A man in the back mutters, "It sounds like a fantasy. How do you
    know AI won't just be as confused as the rest of us?"

    Yoily smiles, almost pityingly. "Because AI doesn't get tired,
    it doesn't get bored, and it doesn't care about the politics. It
    processes every single text, every commentary, every debate. It
    doesn't forget, it doesn't overlook, and it doesn't need to win an
    argument for pride's sake. You ask it a halacha question, and it
    gives you the answer based on all the sources, not just the one that
    the guy down the street likes best. And this already exists.*"

    He looks around, as if daring anyone to argue. The room is silent.

    "And here's the beauty of it," he adds, leaning in. "AI levels
    the playing field. It doesn't matter if you're some HaRav HaGaon
    or just Joe Shmo. You ask a question, and you get the same answer
    as everyone else. Everyone is equal before the machine. Everyone
    has the same access. Democracy in halacha! You think that's not
    revolutionary? Think of it this way. If you've had it with communities
    and Rebbes, and I know, intimately, some people who have, you won't
    need them any more. Because instead you'll have an AI that does
    what they do, only better."

    Someone calls out, "And what if the answer is something we don't
    like?"

    Yoily waves his hand dismissively. "Nu, so we don't like it. Since
    when did truth depend on whether we like it? It's not about comfort;
    it's about keeping halacha real and relevant. Humans aren't capable
    of doing that alone, not anymore. We're too busy bickering over
    whose minhag is older or purer or whatever. AI doesn't have those
    hangups. It just gets the job done."

    He pauses again, his voice lowering, drawing the audience in. He
    focuses on Eden. "By the way, young lady - what was your name? ...yes,
    Ayden - you mentioned a streamlined halacha in the style of the
    Rambam....

    *This really does exist. The way to become a beta tester is to leave
    a comment below briefly stating your thoughts about this series.

RMK heads the Dicta AI lab at Bar Ilan. In this footnote, he is referring
to RabbAI, a project of one of the students there.

Tir'u baTov!
-Micha

-- 
Micha Berger                 Man is equipped with such far-reaching vision,
http://www.aishdas.org/asp   yet the smallest coin can obstruct his view.
Author: Widen Your Tent                            - Rav Yisrael Salanter
- https://amzn.to/2JRxnDF