3 ms·
The whole discussion is about whether or not the second case of using o1 outputs to fine tune R1 is what allowed R1 to become so good. If that's the case then y
by pertymcpert 2y ago
The whole discussion is about whether or not the second case of using o1 outputs to fine tune R1 is what allowed R1 to become so good. If that's the case then your assertion that DeepSeek will surpass OpenAI doesn't really make sense because they're dependent on a frontier model in order to match, not surpass.
- hmottestad 2y agoYeah, that's my point. If they do end up surpassing OpenAI then it would seem likely that they aren't just relying on copying from o1, or whatever model is the frontier model at that time.