4 ms·
Yeah sure, some time in the future. For now I think the fuzzyness and mediocre quality of the responses you get at scale are probably big showstoppers for usage
by theK 3y ago
Yeah sure, some time in the future. For now I think the fuzzyness and mediocre quality of the responses you get at scale are probably big showstoppers for usage at large.
Also, I've not followed this domain over the summer, are we still at 50k token caps on the state of the art models?
- nraford 3y agoClaude has been able to do 100k for months and is as good, if not better (IMHO) than GPT4 for some things.
- theK 3y agoOkay, 100k means we are getting to the point where one can safely edit a large piece of text, say a blogpost or a small scientific paper if you are lucky. It still means most of the long running interactions will get clipped/compressed relatively soon. But good to see the industry evolving