Skip to content

Comment on ChatGPT struggles with a simple logic problem

Comments

I tried to play with this problem (on text-davinci-003, which I prefer to ChatGPT).

It performs considerably worse than it does on many harder problems I gave it before. It seems to capture some architectural weakness in GPT.

Can you provide any details about the harder problems you mention?

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.