News for hackers. Stuff that matters.

A comment on GPT-4: 8 x 220B experts trained with different data/task distributions. Parent comment

Re:GPT-4: 8 x 220B experts trained with different data/task ... by kylewatson on Wednesday June 21, 2023 @11:13AM (#36419526)
I feel like we should try asking it directly. Maybe it would tell us.
[ Parent | on HN ]

There are no comments on this comment yet.