In the past two days, the author has observed at least two other instances where GPT 4's responses were wrong, while GPT 3.5's were not only correct but also of high quality. This has led the author to appreciate the performance of GPT 3.5, similar to their initial experience with GPT 4.
Key takeaways:
- The author has noticed that GPT 3.5 often provides better answers to simple questions than GPT 4.
- They provide examples where GPT 4's answer was incorrect, but GPT 3.5's answer was not only correct, but also of high quality.
- GPT 3.5 was able to anticipate and provide an answer for an order by clause that the author didn't even ask for.
- The author has observed at least two other instances in the past two days where GPT 4's answer was wrong and GPT 3.5's was correct.