Humanist Discussion Group, Vol. 35, No. 478.
        Department of Digital Humanities, University of Cologne
                      Hosted by DH-Cologne
                       www.dhhumanist.org
                Submit to: humanist@dhhumanist.org




        Date: 2022-01-20 13:12:38+00:00
        From: Willard McCarty <willard.mccarty@mccarty.org.uk>
        Subject: intelligence, artificial and biological

To write or speak about artificial intelligence is like attempting to
have a serious conversation in a crowded room of people who are trying
to impress each other, and as more crowd into the room, they talk
louder in order to be heard over the din. You might think it wise to
leave to find a quiet place to think, or leave with someone you happen
luckily to have found who wants a genuine conversation. In plain terms,
instead of talking or writing about AI, you might conclude it's far
better to write and talk about another subject altogether.

Yet to turn your back on this subject, given current beliefs and
technologies, and what is being done with them -- see Stuart Russell's 
recent Reith Lectures and Erik Larson's The Myth of Artificial 
Intelligence (the book is better than its title might suggest) -- is 
perhaps not such a good idea. Do we really want what's being said and 
done to prevail?

Take Alexandre Gefen's upcoming talk, announced on Humanist yesterday,
together with his paper, "AI: a Deep History". (If I may, I suggest that if
the subject interests you, you follow the links, register and download the 
paper.) To be as brief as possible I'll confine myself to his first few 
sentences, which are laid down as simple facts:

> Artificial Intelligence (AI) can be defined as the set of
> mathematical methods and computer technologies designed to solve
> problems ordinarily handled by the human mind, from the accompaniment
> of human tasks (digital tools) to the substitution of humans (the
> horizon of a "general AI" capable of producing reasoning). We are
> currently at an intermediate stage of this technological development,
> which concretely started in the 1950s, but one that have been
> fantasized about for millennia. AI already offers better performances
> than humans for many specialized tasks (image and document
> classification, translation, trend analysis and prediction,
> production of sounds, texts and images, resolution of elaborate
> games). AI has also shown its ability to plausibly simulate certain
> interactions and since 2014 conversational agents have been able to
> pass the Turing test imagined by the mathematician in 1950...

This sounds, I suppose, entirely unremarkable, but on reading it I 
was persuaded that remarks were imperative, especially here, on 
Humanist. But remarking turned into a note far too long, so I will 
confine myself to asking four questions:

(1) Is an imitative AI all that is possible, all that is desirable, all that
has been achieved?

(2) What exactly has humankind "fantasized about for millennia", and how
are these fantasies of the historical past related to current work?

(3) Do "better performances than humans for many specialised tasks" sum
even in principle to intelligence as we strive to understand and develop
it in ourselves?

(4) Have conversational agents in fact passed the Turing test, and if
so, what has this to do with genuine, meaningful conversation between
people?

Here the room is quiet, and everyone is listening. 

Yours,
WM

--
Willard McCarty,
Professor emeritus, King's College London;
Editor, Interdisciplinary Science Reviews;  Humanist
www.mccarty.org.uk


_______________________________________________
Unsubscribe at: http://dhhumanist.org/Restricted
List posts to: humanist@dhhumanist.org
List info and archives at at: http://dhhumanist.org
Listmember interface at: http://dhhumanist.org/Restricted/
Subscribe at: http://dhhumanist.org/membership_form.php