Humanist Discussion Group, Vol. 39, No. 29.
Department of Digital Humanities, University of Cologne
Hosted by DH-Cologne
www.dhhumanist.org
Submit to: humanist@dhhumanist.org
[1] From: Willard McCarty <willard.mccarty@mccarty.org.uk>
Subject: intelligences (16)
[2] From: David Zeitlyn <david.zeitlyn@anthro.ox.ac.uk>
Subject: 39.4: repetition vs intelligence? (13)
[3] From: Gabriel Egan <mail@gabrielegan.com>
Subject: Re: [Humanist] 39.28: repetition vs intelligence (106)
--[1]------------------------------------------------------------------------
Date: 2025-05-27 08:30:07+00:00
From: Willard McCarty <willard.mccarty@mccarty.org.uk>
Subject: intelligences
What's the problem, I wonder, with plural kinds or modes of
intelligence? Why be defensive over other kinds or modes? Are we worried?
I rather like Geoffrey Lloyd's "semantic stretch", for which see
The Revolutions of Wisdom: Studies in the Claims and Practice of
Ancient Greek Science. Sather Classical Lectures, Vol. 52. (University of
California Press, 1987), pp. 175-176. The idea is applied and developed
throughout his many books and papers.
Best,
WM
--
Willard McCarty,
Professor emeritus, King's College London;
Editor, Humanist
www.mccarty.org.uk
--[2]------------------------------------------------------------------------
Date: 2025-05-26 18:46:41+00:00
From: David Zeitlyn <david.zeitlyn@anthro.ox.ac.uk>
Subject: 39.4: repetition vs intelligence?
Hello from Cameroon
I’ve never tried to contribute to humanist from here before
But Tim Smithers talk of “ The word 'intelligence' is what I call an
ice-hockey-puck word.” Prompts me to remind humanist readers of Gallie’s
1956 paper on “essentially contested concepts”. Intelligence is a case
in point. I discuss this and give citations in a 2022 or 2023 article on
humility in JASO Journal of the anthropological society off Oxford. (Oh
the irony of citing myself on humility). This should be easy to find
online if you have good access to tinterweb which I do not from here.
David en brousse
Sent from my iPhone
--[3]------------------------------------------------------------------------
Date: 2025-05-26 11:44:35+00:00
From: Gabriel Egan <mail@gabrielegan.com>
Subject: Re: [Humanist] 39.28: repetition vs intelligence
James Rovira asks two questions
about Word Embedding as performed
by algorithms such as Google's
word2vec:
1) how does the "training process"
identify that words are "close in
meaning"?
The closeness in meaning that we
are referring to here is not the
quality of being synonymous.
The closeness in meaning of the
words 'king' and 'queen' is
a similarity-with-a-distinction,
and that distinction is gender.
That is:
king is to queen
as uncle is to aunt
as rex is to regina
as waiter is to waitress
We wouldn't say that 'uncle' is
a synonym of 'aunt', but they
are close in that both mean
'a sibling of one's parent', but
with a distinction of gender.
In Word Embedding, two words are
identified as being close in meaning
(in the above sense) by their both
being found in the company of the
same set of other words ('the company
they keep'). This is not, as Tim
Smithers has it, the same as saying
that words are similar in meaning
if they "frequently occur close
to each other". Rather, two words
are similar in meaning if they
share a high likelihood of being
seen with the same set of other
words near them.
That is what I meant by "the
things we say of kings are
like the things we say of
queens". These words are
similar because they are
more than usually likely
to have near them words such
as 'sovereign', 'reign',
'monarch', 'throne' 'crown',
'usurp', 'abdicate', and
so on.*
James's second question is:
2) "And how would the process
that you describe generate
readable text rather than,
say, a list of words resembling
a thesaurus?"
Answer: why would it do that? A
thesaurus gathers words that
are synonyms (and sometimes
antonyms). It doesn't capture
the similarity of meaning
discussed above.
Regards
Gabriel Egan
* Of course, there is another
set of words that also are
likely to be found in the
company of 'king' and 'queen',
and they are 'pawn', 'rook',
'bishop', 'knight', 'mate',
'check', and so on.
One or more of the numbers
in the vector for 'bishop'
will capture its shared
meaning with other chess
pieces, and another one
or more of the numbers
in the vector for 'bishop'
will also capture its shared
meaning with terms such
as 'priest', 'prelate',
'deacon', 'vicar', and so
on. That is, words are
similar to one another
along multiple dimensions
at once. Large Language
Models became impressive
as Artificial Intelligence
only once we gave them
enough dimensions to capture
something of how we carve
up reality ourselves. This
really is, pace Smithers,
all about semantics.
_______________________________________________
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
