logoalt Hacker News

geraneum • yesterday at 11:37 PM • 1 reply • view on HN

They're trained on all content they could access produced by humans. The op is asking for the most probable string of token extracted from that content following the prompt.


Replies