That's not widely true. E.g the GPT 4 tech report pointed out nearly all their e... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		jph00 7 months ago \| parent \| context \| favorite \| on: What's the strongest AI model you can train on a l... That's not widely true. E.g the GPT 4 tech report pointed out nearly all their experiments were done on models 1000x smaller than the final model.

tmule 7 months ago [–]

Fair point, though I’d argue that there’s inherent selection bias for improvements that could fit a scaling law curve in the small model regime here.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact