Humanist Discussion Group, Vol. 36, No. 553.
Department of Digital Humanities, University of Cologne
Hosted by DH-Cologne
www.dhhumanist.org
Submit to: humanist@dhhumanist.org
Date: 2023-04-26 10:59:25+00:00
From: Manfred Thaller <manfred.thaller@uni-koeln.de>
Subject: Some naive musings about AI / artificial intelligence
Dear Willard, Dear all,
I am most certainly not a specialist of AI or artificial intelligence.
Nevertheless, recent discussions, even more so outside of Humanist than
within, tempt me to share a few naive observations. As these touch upon
a variety of topics, I’d like to present them as a lose set of theses.
(1) Some years / decades ago most discussions of AI / artificial
intelligence usually started with explaining the difference between a
“strong paradigm of AI” and a “weak paradigm of AI”. Wikipedia confirms
my impression that “The differences between weak vs. strong AI is [sic]
not widely catalogued out there at the moment.”
So, just as a reminder:
The strong paradigm of AI / artificial intelligence requires a computer
system to “implement a software system that thinks like a human being”.
The weak paradigm of AI / artificial intelligence requires a computer
system to “create the impression of a software system that thinks like a
human being”. (Trusting Wikipedia that there are currently no fully
orthodox definitions I quote the ones I myself used in teaching.)
While I am afraid the discussion about AI / artificial intelligence
outside of Humanist may be a bit too excited momentarily to expect clean
terminologies, I have the impression that colloquially “AI” is generally
used to speak about industrial applications (to the best of my knowledge
100% weak paradigm) while “artificial intelligence” tends to refer to
speculations about the strong paradigm (as far as I know not realized
anywhere yet).
Now the weak paradigm can do astonishing things, as many challenges
which allegedly require a strong solution would probably overtax a
strong solution in any case just as well as a weak one. In Germany,
e.g., the discussion about autonomously driving cars was for some time
enriched by the argument, that no steering AI could be relied upon to
make in a split second the ethically correct decision in a situation
where only two steering maneuvers would be possible: one risking to kill
an elderly person in a wheel chair on one side of the street, the other
killing a toddler on the other side. As we all probably agree that each
and every human driver would make this decision ethically correct in a
split second … in the real world we can go far within the weak paradigm.
(2) Two years ago, discussions of artificial intelligence automatically
flowed into autonomous driving; today this automatism has shifted
towards ChatGPT.
For me, the truly interesting thing about it is, how fascinatingly
simple the conceptual model is: if you check which word most frequently
follows another in a great body of sentences, that sequence most
probably will make sense again. Let my emphasize, very much so, that
calling this “fascinatingly simple” should in no way reduce the great
respect for the people who realized the “simple matter of programming”
to turn this into a working system. Kudos; almost proskynesis would be
an appropriate appreciation for the engineering.
Indeed, like everybody else, I’ve been deeply impressed by the beautiful
flow of the language generated. This seems to be (= weak paradigm) a
very good implementation of seeming creativity.
Less impressive the content generated: Today the “Write a biography of
Manfred Thaller” informs me that I have been a major contributor to and
propagator of TEI and CIDOC-CRM, which is a surprise not only to me, and
reminds me of my time as director of the Bavarian State Library in
Munich, as well as of the reception of the Roberto Busa Prize in 2013.
The only sad thing is, that I already died in 2018.
Asking in German yesterday left me alive and withheld the Busa Prize and
the directorship; memberships in the Norwegian Academy of Science and
the Academy of Science and Literatur in Mainz were some compensation. My
wife was very pleased to be reminded in her biography of the many world
records she had set and the gold medals she received at various Olympics
in breaststroke swimming.
This seems to be a very good implementation of creativity; facticity:
not so good.
(3) How creative can a Human be?
If I write the sentence “Wagrumble stromblebum sequat alomeny.” I am
highly creative. In all probability during the one million years of
humanoid existence no other humanoid has uttered exactly that sentence.
Unfortunately no other humanoid has a chance to understand it, as a side
effect.
/Ash nazg durbatulûk,//
//ash nazg gimbatul,//
//Ash nazg thrakatulûk//
//agh burzum-ishi krimpatul.//
///
at first look seems to be just as creative. At second this is not so –
consult J.R.R. Tolkien who wrote the sentence before me. Well; quoting
Wikipedia a second time: the <emph>text</emph> is fully creative; the
<emph>concepts</emph> behind it are traced by some into Norse mythology,
Wagner, Roman religious practice and Plato’s Republic. Which may explain
partially, why the text is acceptable.
(Probably quite un-creative) thesis: Creativity is the art to make the
largest possible change in the configuration of agreed upon semantic
symbols without rendering them incomprehensible.
Corollary: If that is so, creativity can be achieved by the weak paradigm.
(4) Completely human, anecdotal and deplorably cynic comment: That focus
on a limited amount of change explains a lot, particularly in the
development of and the application of information technology.
“Creativity” in interface design and the shifting popularity of specific
computer applications in the Humanities always lead me to wonder whether
humanoids might have inherited one or two genes from lemmus lemmus.
(a.k.a.: lemming)
(5) Unfortunately, as my unknown academic achievements and my sad demise
in 2018 quoted above show, connecting a successful re-configuration of
semantically loaded symbols MIGHT create well readable literature (I
confess to being partial to SF and fantasy), but at the current stage of
the art not to a text confined by the physical world (I swear I am
currently still alive) and the facticity of socio-cultural-economic
reference system (I have definitely never been director of the Bavarian
State Library, nor received the honors quoted).
Any corollaries for literary studies are out of my domain. As
non-student of literature: Could successful literary works be successful
bets, how much creativity the readers are willing to stomach?
This seems to imply, that creativity – or at least a large part of it –
might to be possible within the weak paradigm of AI / artificial
intelligence, while the “openness to the world, and hence to Being”
could require more. (I grudgingly admit the reference to Heidegger,
which should not be construed as any indication of general respect for him.)
This, of course, seems to turn around the almost general consent:
Creativity is allegedly a human privilege; the creation of
interconnections between factlets can be left up to software.
Now this allegation may require, nay, certainly requires, additional
discussion, as the connection between semantically loaded symbols and
factlets in reality are less cleanly separated from each other than
assumed by the previous sentence.
(6) In at least some of the Humanities, however, one could wonder,
whether that might not imply a shift of emphasis. In history – always on
the top of my mind, when I write about “the Humanities” –: could that
mean that in an extreme view, all the Hayden White parts (“writing of
history is literature”) could be safely left to ChatGPT 7.0 while the
human historian is required to reconstruct the most appropriate
interconnections of the factlets?
Where, of course these interconnections are much better controlled, if
they are handled by forms of (computer supported) publication as derived
from Jean Claude Gardin’s logicist analysis, claiming the “possibility
to reconstruct scientific reasoning in the human sciences in general and
in archaeology in particular”.
( Jean-Claude Gardin: Une archéologie théorique, Hachette, 1979;
Jean-Claude Gardin: Le calcul et la raison: essais sur la formalisation
du discours savant, Éd. de l’École des hautes études en sciences
sociales, 1991, 65.
More easily accessible - and downloadable - in English:
https://www.researchgate.net/publication/242097181_On_Jean_Claude_Gardin's_logic
ist_analysis
; https://zenodo.org/record/6498979/files/bbaw-talk.pdf?download=1 )
One should point out, that Gardin’s thinking was an integral part of one
of the earlier boom phases in the traditional boom – bust cycle of AI /
artificial intelligence. (The “expert system” hysteria in his case.)
Whether the support of such a form of publication is covered by the weak
paradigm or requires the strong remains to be seen.
(7) (At least if you are a historian) If you are afraid, that a weak
paradigm tool like ChatGPT is able to submit a valid student’s paper:
maybe you should change the assignments?
Apologies for being loquacious,
Manfred
--
Prof.em.Dr. Manfred Thaller
formerly University at Cologne /
zuletzt Universität zu Köln
_______________________________________________
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
