AI/ML(Machine learning)

Some thoughts on AI/ML(Machine learning) and a psak generator:
But it will happen imho. And since lots of psak (if not all) has lots of unarticulated premises, it will be interesting to see what ML extrapolates. In the outside world the man machine partnership is one model (others say machine only!) Any thoughts on how this might play out?
KT
Joel Rich

THIS MESSAGE IS INTENDED ONLY FOR THE USE OF THE
ADDRESSEE. IT MAY CONTAIN PRIVILEGED OR CONFIDENTIAL
INFORMATION THAT IS EXEMPT FROM DISCLOSURE. Dissemination,
distribution or copying of this message by anyone other than the addressee is
strictly prohibited. If you received this message in error, please notify us
immediately by replying: "Received in error" and delete the message.
Thank you.

In terms of the tech, we're not close yet, and may never be.

To weave a derashah into my explanation of that sentence, the word
"binah" can refer to thought in at least two ways:

- from "bein", the art of making distinctions -- bein qodesh lekhol, bein
  or lechoshekh... Being able to characterize this as an A, that as a B,
  this is spam, this is not spam. And when things are not separated:
  this A and B share features that would allow them to belong to the same
  font, whereas that A & B are in another.
  
  Finding the limits of sets, to know what to separate. That's what AI is
  now doing. Using what boils down to a lot of linear algebra to draw lines
  between sets using known examples of what's in and what's out.

  That can be obvious in cases of AI reading CT or MRI scans. But that is
  also how automous vehicles decide when to change lanes, accelerate,
  or whatnot. There are two sets: a right time and a wrong time.

- from "binyan", the art of combining facts. Whether making deductions of
  the sort you did formally in Geomatry class, or inductions -- figuring
  out that those sets even exist.

  There is AI that does binyan-binah. It is very simple, generally just
  searching the space of possibilities until it hits sueccess (as defined
  for it). That's not what ML is about, and it's not where we are close to
  the commercial application. And may never be.

Then there is informal thought. Conclusions you can only reach because
you not only have a concept of red in your head, but you know what it
is like to see red. (Apologies to the color blind for that example.)

In terms of hashkafah...

I argued that lo bashamayim hee has more to do with whose hands halakhh
is in than whose it isn't. A pesaq can only come from someone who (to
jump ahead to the end of the statement) "befikha ublivavkha la'asoso".

Related to that knowing what red is *like*, knowing what moral is *like*
is arguably a fundamental element of what a poseiq brings to the question.

And even if that argument fails hashkafically, halachically pro forma a
poseiq must be a mentally competent Jewish male adult who you believed
has the appropriate knowledge, ability and experience (experiential
knowledge?). It's the difference between whether any error would be your
aveira beshogeig or your doing the right thing in following a pesaq. (And
the the case of a human poseiq, their shogeig.)

Tir'u baTov!
-Micha

Some thoughts on AI/ML(Machine learning) and a psak generator:
But it will happen imho. And since lots of psak (if not all) has lots of
unarticulated premises, it will be interesting to see what ML extrapolates.
In the outside world the man machine partnership is one model (others say
machine only!) Any thoughts on how this might play out?

I can't recall if somebody already sent a link to this interesting article
(to me: utterly fascinating), which is davka on AI/ML vis-a-vis texts. In
it, he talks about future possibilities.

https://thelehrhaus.com/commentary/torah-study-and-the-digital-revolution-a-glimpse-of-the-future/

(PS: some of the future tools are available on-line, free, in beta form
already.)

It seems to me that AI/ML might *help* a posek, in discovering potentially
unarticulated premises -- but the posek himself would need to judge whether
the machine-made connections make sense or not.

-- Sholom