Hacker News new | ask | show | jobs
by michaelhartm 1177 days ago
They used the 6b GPT4-J, not 20B. That's what's interesting, it's a smallish large language model :).
1 comments

GPT-J, not GPT4-J.