Humanist Discussion Group, Vol. 35, No. 567.
Department of Digital Humanities, University of Cologne
Hosted by DH-Cologne
www.dhhumanist.org
Submit to: humanist@dhhumanist.org
[1] From: Simon Rae <simon.rae@gmail.com>
Subject: Re: [Humanist] 35.566: consider ELIZA: what do we have that Weizenbaum didn't? (74)
[2] From: Gioele Barabucci <gioele.barabucci@ntnu.no>
Subject: Re: [Humanist] 35.566: consider ELIZA: what do we have that Weizenbaum didn't? (33)
--[1]------------------------------------------------------------------------
Date: 2022-03-03 13:34:38+00:00
From: Simon Rae <simon.rae@gmail.com>
Subject: Re: [Humanist] 35.566: consider ELIZA: what do we have that Weizenbaum didn't?
Hi Willard,
Just off the top of my head, these days if one were to repeat the ELIZA
experiment, the big difference I suspect would be a sense of place. I'm
guessing that the original ELIZA existed on a large mainframe computer
where those taking part in the conversation were very aware that they were
'talking' to a machine. Nowadays, presumably, the experiment would be done
using smartphones and participants would not know they were doing anything
different than their normal phone chat. The researchers might indeed
introduce some elements of normality - brief loss of signal, traffic noises
etc as a way to make any conversation with ELIZA Mark 2 seem totally
normal. Judging by many overheard conversations on the train I suspect that
many would be completely fooled by anything that appeared to conform to
what they wanted, or expected to hear. I think that the user interface
would be the biggest difference.
I do recall reading, at about the same time of ELIZA Mark 1, of work done
by Christopher Strachey on using a teletype machine and a simple (!)
program to ask preliminary questions of patients in a Doctor's GP
surgery (I think, it's a while ago and I didn't make a note of the
reference, mea culpa). It appeared, it seemed, that people were quite happy
to input personal and sometimes embarrassing details to the program
concerning their ailment, happier indeed than answering the same questions
when asked by the receptionist. Sometimes, perhaps, a cold emotionless
interface can be a bonus, just so long as it asks sensible questions.
Cheers
Simon
Simon Rae
Retired.
On Thu, Mar 3, 2022 at 8:37 AM Humanist <humanist@dhhumanist.org> wrote:
>
> Humanist Discussion Group, Vol. 35, No. 566.
> Department of Digital Humanities, University of Cologne
> Hosted by DH-Cologne
> www.dhhumanist.org
> Submit to: humanist@dhhumanist.org
>
>
>
>
> Date: 2022-03-03 08:25:07+00:00
> From: Willard McCarty <willard.mccarty@mccarty.org.uk>
> Subject: consider ELIZA
>
> Just to open up a discussion of where we're at technologically speaking,
> consider Weizenbaum's ELIZA, written to demonstrate the triviality of
> engaging people in conversation with machines. Its wild success, which
> he thought showed disturbing delusional thinking, did esrablish
> something important, I think: the appeal of therapeutic conversation
> with a machine.
>
> We're still at it, of course. My question is the simple one I've already
> asked. It requires a considerably greater breadth of knowledge than I
> have. So, I ask again (with apologies): what do we have to hand that
> he didn't? Can we say simply 'better hardware' and be done with that
> side of it, or are there advances in hardware beyond speed and mnemonic
> capacity that need to be mentioned? If we were talking to him now what
> would we have to explain?
>
> Many thanks for answers and suggestions.
>
> Yours,
> WM
> --
> Willard McCarty,
> Professor emeritus, King's College London;
> Editor, Interdisciplinary Science Reviews; Humanist
> www.mccarty.org.uk
--[2]------------------------------------------------------------------------
Date: 2022-03-03 15:07:10+00:00
From: Gioele Barabucci <gioele.barabucci@ntnu.no>
Subject: Re: [Humanist] 35.566: consider ELIZA: what do we have that Weizenbaum didn't?
Dear Willard,
here are a few meaningful differences that I can think of:
1) _Much_ better hardware. In 1966 the amount of available working
memory and storage was counted in bytes. Just adding a new phrase to
Eliza may have required removing some of the code that drove the logic.
2) More digitized data. In the past 60 years humanity has accumulated a
wealth of data in digital form and semi-structured, in particular
conversations between people, that was simply not available when Eliza
has been created.
3) Better statistical methods. Aside from all the machine-learning hype,
there are now much better mathematical and statistical methods for the
analysis and computation of model-free inference, non-parametric
statistics and for information retrieval (e.g., SVM, Fisher kernel).
4) Advancements in theoretical information science. Fundamental things
like Earley parsers, PEGs, or polynomial algorithms for linear
programming (i.e., finding "the best element in a set" under set of
constraints) were yet to be discovered.
Combined together, all these improvements would probably give rise to a
much different Eliza today. Or maybe not, if we assume that Eliza's main
point was to point out the quirks of human understanding and empathy.
Regards,
--
Prof. Dr. Gioele Barabucci <gioele.barabucci@ntnu.no>
Associate Professor of Computer Science
NTNU — Norwegian University of Science and Technology
_______________________________________________
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
