4 ms·
> Is it possible that GPT-4 already had this new optimization somewhere in its training set? Isn't it necessary for GPT-4 to have had the optimization somewher
by tejohnso 3y ago
> Is it possible that GPT-4 already had this new optimization somewhere in its training set?
Isn't it necessary for GPT-4 to have had the optimization somewhere in its training set given how it works?
- wilg 3y agoNo, it's not necessary for it to have had the optimization somewhere in its training set given how it works. LLMs can solve problems they were not trained on, with varying degrees of success.