Skip to content

Comment on GPT-4-turbo preliminary benchmark results on code-editingparent

Comments

I understand your broader point, which is that progress in this space will be discontinuous and hard to see until it’s already happened.

But two years ago, GPT-3 was already astounding. ChatGPT made some pretty huge gains, and GPT4 more still, but the thing that exists today was relatively predictable, compared to a world where human effort is no longer important.

AboutSource Built by g1lg1l

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