News for hackers. Stuff that matters.

A comment on GPT-4. Parent comment

Re:GPT-4 by baq on Tuesday March 14, 2023 @04:33PM (#35158039)
The problem with this simplification is a bog standard Markov chain fits the description as well, but quality of predictions is rather different.

Yes the LLM does generate text. No it doesn’t ‘just generate text that’s it’.

[ Parent | on HN ]
Re:GPT-4 by peterashford on Tuesday March 14, 2023 @06:14PM (#35159778)
"It generates text better than a Markov chain" - problem solved
[ Parent | on HN ]
Re:GPT-4 by baq on Wednesday March 15, 2023 @02:53AM (#35164793)
Classic goal post moving.
[ Parent | on HN ]
Re:GPT-4 by peterashford on Thursday March 16, 2023 @06:51PM (#35190034)
Not really, I think the original post was just being a post, not a scientific paper. Sometimes people speak normally
[ Parent | on HN ]
Re:GPT-4 by Izkata on Wednesday March 15, 2023 @05:23AM (#35165736)
The biggest problem I've seen when people try to explain it is in the other direction, not people describing something generic that can be interpreted as a Markov chain, they're actually describing a Markov chain without realizing it. Literally "it predicts word-by-word using the most likely next word".
[ Parent | on HN ]