[Avodah] ai

[Micha Berger]

On Wed, May 17, 2023 at 05:48:11AM +0300, Joel Rich via Avodah wrote:
> There is certainly a permanent spiritual argument that artificial
> intelligence could never becoming a poseik...

Since there are strong arguments for saying a woman can't, lo kol shekein.

>                 However, the arguments of not having shimush or a mesora
> seem to me to be more short-term practical arguments. They seem to be based
> on the fact that AI has no access to this data because it is not part of
> the written corpus....

I think it gets to something else -- the difference between data and
wisdom. Or, as you put it:

> It seems this is a subset of the great debate in the outside world
> concerning creativity, the hard problem of consciousness (and what it's
> like to be a bat), the existence of free will, and can AI have
> consciousness without a body or be sentient. (The creativity issue is an
> interesting one, because from what I understand, alpha go revealed
> strategies which humans hadn't discovered).

None of which relates to what currently exists and is called AI today.
Today's "AI" programs are artificially producing results we normally
associate with intelligence without trying to produce anything like
a machine with free will or consciousness / first-hand experiences.

Simulated intelligence.

AI in the sense of synthetic intelligence is still Science Fiction,
and may not be possible. People may not be able to understand minds
well enough to make one. Or, while the brain is a "shadow" of the soul
in the physical world, you may need an actual soul to make a brain
that isn't a philophical "zombie". (I.e. something that acts exactly
like a person does, but without an "I", no first-hand experiences.
Assuming such a thing is possible.)

This distinction between simulated and synthetic intelligence is my own.
But this marketing of neural nets, a bunch of linear algebra, as "AI",
requires we re-introduce clarity to the topic somehow.

When computer chess playing stopped being amazing, we stopped calling it
AI. The same may happen here. Or it may not, given how much more central
to the human experience text communication is.

Some of how we assess AI is pareidolia. The human tendency to see a face
in things like a car's headlights and bumper. We are built to overestimate
how human something looks.

But still, it is incredible how intelligent something can seem when
it isn't even built to model the topic under discussion. Chat GPT is a
"language model"; it produces texts that make sense by analyzing texts.

It is closer to the predictive text in a Google search window than
intelligence.

Personally, I don't think one can even get to AGI -- artificial general
intelligence. A perfect simulated intelligence, a philosphical zomie.
(Never mind a synthetic intelligence.) After all, a soul is tzelem
Elokim. There is something not only unique about the human soul, it has
a unique image.

Tir'u baTov!
-Micha

PS: I use chatGPT backwards. Given that it is designed to produce texts
typical of the internet, I use its results as an indicator of where human
discourse is. For example, without specific prompting, its "devar Torah on
parashas X" will not quote Chazal or Rishonim. It will give self-help advice
or self-evident feel-good material (glurge) directly off the text. I am
guessing that speaks volumes about what it being put on-line about the
parashah. (Although it is likely that most such pages are not from Orthodox
authors.)

-- 
Micha Berger                 Today is the 48th day, which is
http://www.aishdas.org/asp   6 weeks and 6 days in/toward the omer.
Author: Widen Your Tent      Yesod sheb'Malchus: What binds different
- https://amzn.to/2JRxnDF           people together into one cohesive whole?