LLMs can't self-correct in reasoning tasks, DeepMind study finds

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by bondarchuk on Monday October 09, 2023 @03:59PM (#37824747)

There used to be 2 garbagemen on every garbage truck in my city, now they have a remote-controlled crane which is operated by a single garbageman who also drives the truck and presumably they fired the other. Garbageman 2x productive > 2x fewer garbagemen. This is a concrete example that contradicts your claim.

[ Parent | on HN ]

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by astrange on Monday October 09, 2023 @04:20PM (#37825001)

Or they didn't fire them and they're doing another task in the same job. Or they're on another route and they run more garbage trucks than they used to, since population has grown.

[ Parent | on HN ]

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by hewlett on Monday October 09, 2023 @04:59PM (#37825414)

I doubt they were fired. sanitation workers have pretty strong unions

[ Parent | on HN ]