4 ms·
I use GPT 4, and it still constantly invents things and presents them to me with authority. I haven’t tried GPT-4o yet.
by camdenreslink 2y ago
I use GPT 4, and it still constantly invents things and presents them to me with authority. I haven’t tried GPT-4o yet.
- nopinsight 2y agoI do find that current LLMs are quite bad at design problems and answering very specific questions for which they may lack sufficient training data. I like them for general Q&A though. A different architecture or an additional component might be needed for them to generalize better for out-of-training-distribution questions.