[seqfan] Re: Sequence from ChatGPT

[Frank Adams-watters]

I.e., use the "less" keyword.

Franklin T. Adams-Watters


-----Original Message-----
From: William Keith <william.keith at gmail.com>
To: Sequence Fanatics Discussion list <seqfan at list.seqfan.eu>
Sent: Sun, Apr 2, 2023 10:33 am
Subject: [seqfan] Re: Sequence from ChatGPT

No, it is not an author.  You got an idea for a perfectly good
sequence by watching a Markov chain generate random words (to which
you had to add mathematical precision anyway).

The fact that GPT blithely appends utter nonsense as both numerical
example and textual explanation should be warning enough that in its
current state it is an utterly inappropriate tool for mathematical
investigation, and I would do nothing to encourage its use.

As the very *first* example of the use of a language model as part of
the process of generating a sequence for the OEIS there is perhaps
some minor historical interest, which could go in the comments.  I
would in fact add the editorial comment that sequences generated with
the aid of such models are likely to be looked upon unfavorably as
contributions without further evidence of independent mathematical
interest in the sequence.

William

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