Skip to content

Comment on It's not just statistics: GPT-4 does reasonparent

Comments

The author wrote a thoughtful article attempting to break this down and has humbly popped into the comments to discuss it... Is your direct response really just to cross your arms and say nope? Like, really?

Using the metric “can reason” on a LLM is like using the metric “can bleed” on a stone.

Maybe some red stuff comes out when you break it. Is it blood, or is it something pumped into the other side?

AboutSource Built by g1lg1l

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