[seqfan] Re: ChatGPT on seqfan

[Ali Sada]

 I think the sequence could be added to the OEIS for historical reasons, as William Keith said. If I remember well, ChatGPT has failed previously to produce a correct OEIS sequence. The OEIS might be recording a milestone in AI history. It might not be mathematically interesting now, but maybe Robert, or someone else, could try again in two or three years and see how AI developed in this regard. 

Best,
Ali 


    On Monday, April 3, 2023 at 04:37:40 AM GMT+1, Olivier Gerard <olivier.gerard at gmail.com> wrote:  
 
 Thanks to all of those who reacted to Robert Israel's anecdote as well as
other inputs.

There are already very interesting conversations going on (for instance on
Scott Aaronson's blog that
has been quoted in this thread). Scott is currently working with OpenAI and
has always
had very strong opinions on these subjects.
Let's transfer discussions about AI and mathematics in these kinds of
forums.

Curiosity is good and fun. I enjoy coercing ChatGPT into generating correct
answers
(especially code) and I learn some things along the way.
There are certainly ways we can use ChatGPT and similar
tools as assistant for Encyclopedia's activity or related research, but as
others and I
have already said several times, we do not want to encourage spam and random
applications for new sequences, especially by people who cannot even
simulate
the artificial or second-hand knowledge of ChatGPT.

One way to enhance all these systems is to publish more human language
essays
about important sequences and general principles of combinatorics and its
relation to other topics. Another way is to write extensively on common
mistakes
and misconceptions in the field.

Members are welcome to signal, index, translate or extend such resources as
they discover them.


Olivier GERARD
Seqfan Mailing List Administrator

--
Seqfan Mailing list - http://list.seqfan.eu/