Path: ...!news.nobody.at!weretis.net!feeder8.news.weretis.net!reader5.news.weretis.net!news.solani.org!.POSTED!not-for-mail
From: Mild Shock <janburse@fastmail.fm>
Newsgroups: comp.lang.prolog
Subject: 2nd Cognitive Turn ~~> no Bayesian Brain (Re: Prolegomena by
 Rappaport)
Date: Sat, 3 Aug 2024 22:50:14 +0200
Message-ID: <v8m566$ofoq$1@solani.org>
References: <b406aa35-c39b-46f3-862f-1cc4b75143ae@googlegroups.com>
 <1b7ce2bd-722b-4c2e-b853-12fc2232752bn@googlegroups.com>
 <v6731e$6eft$1@solani.org> <v8e611$k6gf$2@solani.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Injection-Date: Sat, 3 Aug 2024 20:50:14 -0000 (UTC)
Injection-Info: solani.org;
	logging-data="802586"; mail-complaints-to="abuse@news.solani.org"
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Firefox/91.0 SeaMonkey/2.53.18.2
Cancel-Lock: sha1:GV3IFTX7x4hQi20B4/EV9KEUAls=
In-Reply-To: <v8e611$k6gf$2@solani.org>
X-User-ID: eJwFwYEBwDAEBMCV8E/ScRD2H6F3jtDow/Cgry9z9t1UxICt707BtmQp3kgbs1OipHbOR7h+I+ktrch68wNlTxXz
Bytes: 5621
Lines: 108

Hi,

Yes, maybe we are just before a kind
of 2nd Cognitive Turn. The first Cognitive
Turn is characterized as:

 > The cognitive revolution was an intellectual
 > movement that began in the 1950s as an
 > interdisciplinary study of the mind and its
 > processes, from which emerged a new
 > field known as cognitive science.
https://en.wikipedia.org/wiki/Cognitive_revolution

The current mainstream believe is that
Chat Bots and the progress in AI is mainly
based on "Machine Learning", whereas

most of the progress is more based on
"Deep Learning". But I am also sceptical
about "Deep Learning" in the end a frequentist

is again lurking. In the worst case the
no Bayension Brain shock will come with a
Technological singularity in that the current

short inferencing of LLMs is enhanced by
some long inferencing, like here:

A week ago, I posted that I was cooking a
logical reasoning benchmark as a side project.
Now it's finally ready! Introducing 🦓 𝙕𝙚𝙗𝙧𝙖𝙇𝙤𝙜𝙞𝙘,
designed for evaluating LLMs with Logic Puzzles.
https://x.com/billyuchenlin/status/1814254565128335705

making it possible not to excell by LLMs
in such puzzles, but to advance to more
elaborate scientific models, that can somehow

overcome fallacies such as:
- Kochen Specker Paradox, some fallacies
   caused by averaging?
- Gluts and Gaps in Bayesian Reasoning,
   some fallacies by consistency assumptions?
- What else?

So on quiet paws AI might become the new overlord
of science which we will happily depend on.

Jeff Barnett schrieb:
 > You are surprised; I am saddened. Not only have
we lost contact with the primary studies of knowledge
and reasoning, we have also lost contact with the
studies of methods and motivation. Psychology
was the basic home room of Alan Newell and many
other AI all stars. What is now called AI, I think
incorrectly, is just ways of exercising large amounts
of very cheap computer power to calculate approximates
to correlations and other statistical approximations.

The problem with all of this in my mind, is that we
learn nothing about the capturing of knowledge, what
it is, or how it is used. Both logic and heuristic reasoning
are needed and we certainly believe that intelligence is
not measured by its ability to discover "truth" or its
infallibly consistent results. Newton's thought process
was pure genius but known to produce fallacious results
when you know what Einstein knew at a later time.

I remember reading Ted Shortliffe's dissertation about
MYCIN (an early AI medical consultant for diagnosing
blood-borne infectious diseases) where I learned about
one use of the term "staff disease", or just "staff" for short.
In patient care areas there always seems to be an in-
house infection that changes over time. It changes
because sick patients brought into the area contribute
whatever is making them sick in the first place. In the
second place there is rapid mutations driven by all sorts
of factors present in hospital-like environments. The
result is that the local staff is varying, literally, minute
by minute. In a days time, the samples you took are
no longer valid, i.e., their day old cultures may be
meaningless. The underlying mathematical problem is
that probability theory doesn't really have the tools to
make predictions when the basic probabilities are
changing faster than observations can be
turned into inferences.

Why do I mention the problems of unstable probabilities
here? Because new AI uses fancy ideas of correlation
to simulate probabilistic inference, e.g., Bayesian inference.
Since actual probabilities may not exist in any meaningful
ways, the simulations are often based on air.

A hallmark of excellent human reasoning is the ability to
explain how we arrived at our conclusions. We are also
able to repair our inner models when we are in error if
we can understand why. The abilities to explain and
repair are fundamental to excellence of thought processes.
By the way, I'm not claiming that all humans or I have theses
reflective abilities. Those who do are few and far between.
However, any AI that doesn't have some of these
capabilities isn't very interesting.

For more on reasons why logic and truth are only part of human
ability to reasonably reason, see

https://www.yahoo.com/news/opinion-want-convince-conspiracy-theory-100258277.html

    -- Jeff Barnett