Humanist Discussion Group

Humanist Archives: Feb. 15, 2022, 7:18 a.m. Humanist 35.529 - GPT-3 and generated poetry

              Humanist Discussion Group, Vol. 35, No. 529.
        Department of Digital Humanities, University of Cologne
                      Hosted by DH-Cologne
                Submit to:

    [1]    From: James Rovira <>
           Subject: Re: [Humanist] 35.526: GPT-3 and generated poetry (30)

    [2]    From: Willard McCarty <>
           Subject: GPT-3 (or similar) and generated conversation? (41)

        Date: 2022-02-14 15:02:41+00:00
        From: James Rovira <>
        Subject: Re: [Humanist] 35.526: GPT-3 and generated poetry

I think if there's one kind of poem a computer will always be able to
generate, it's a cento :).

Jim R

On Mon, Feb 14, 2022 at 1:37 AM Humanist <> wrote:

> I do not know what I am saying
> The world is spinning round
> I am dizzy and I cannot see
> I want to sleep
> And when I wake up
> I will see things more clearly
> IMHO this is a pretty good summary of Tzara's poetry.
> I'm not sure what to make of this. Thoughts?
> mw
> --
> Mark B. Wolff, Ph.D.
> Professor of French
> Chair, Modern Languages
> One Hartwick Drive
> Hartwick College
> Oneonta, NY  13820
> (607) 431-4615

        Date: 2022-02-14 07:05:10+00:00
        From: Willard McCarty <>
        Subject: GPT-3 (or similar) and generated conversation?

This in response to Mark Wolff's question of yesterday. So far, as far
as I know, the trials have been one-off performances in which someone
sends something to GPT-3, something comes back, and we all marvel,
grumble, comment etc. Has anyone tried anything conversational with such
an agent? Interchanges with ELIZA and the like depended on close
imitation of what one might expect another person, in that case someone
in the manner of a therapist, to say, and naive acceptance by the human
who initiates. What if the human using GPT-3 were to attempt a
conversational exchange?

Some years ago a student at Berkeley, Dunbar Aitkens, invented a game
inspired by Hermann Hesse's Glasperlenspiel, which he and his friends
called The Glass Plate Game. See <> for the
current (fossilised?) state of his project. It uses images on cards to
induce conversation among people in a group. I have not plumbed the
depth of the game, but I can see how it might serve to call up hidden
thoughts and develop them. One can imagine a computational -- 
combinatorial -- implementation.

For something along the lines of a GPT-3 conversation or development of
The Glass Plate Game to prove useful, perhaps therapeutic, wouldn't the
problem be to find the 'sweet spot' between mirrored responses and
utterly chaotic ones? The "edge of chaos" in complexity theory comes to
mind. This figure of speech unfortunately suggests something like an 
external border one close to, but the 'space' in which interesting things 
happen is far more interesting than that.

For me the significant point in all this is reached when the interchange
becomes a conversation that unexpectedly provides insight for the human
participant(s). The emphasis, it seems to me, needs to be on the
conversation, not the mechanism, whether this be GPT-3 or a bunch of cards.


Willard McCarty,
Professor emeritus, King's College London;
Editor, Interdisciplinary Science Reviews;  Humanist

Unsubscribe at:
List posts to:
List info and archives at at:
Listmember interface at:
Subscribe at: