4 ms·
To solve it doesn't mean we have to eliminate it completely. I think GPT has solved it to enough extent that it is reliable. You can't get it to easily hallucin
by simianwords 8mo ago
To solve it doesn't mean we have to eliminate it completely. I think GPT has solved it to enough extent that it is reliable. You can't get it to easily hallucinate.
- disgruntledphd2 8mo agoIt depends on how much context is in the training data. I find that they make stuff up more in places where there isn't enough context (so more often in internal $work stuff).