Hacker News Reader

Did you try just once? https://chat.openai.com/share/fa01b895-2b4e-47c1-bbbb-1cef0a...

Perhaps your past discussion, custom prompt or chat temperature (that cannot be adjusted except in the API) caused that to happen. It's not a miracle maker but a stochastic tool.

View on HN

Yes, once. It was the first prompt like "What is ...?". The second prompt was like "Suppose you are a 20 Nobel prize winner in physics etc", and it failed after the second one also.

If you asked them in the same session, the previous answer would have affected the next answer.

The rule of thumb goes: never argue with an LLM, unless you are looking for an argument. Just reset and start over (or edit the prompt, which is the same thing but preserves the previous communication).

The first prompt was actually more specific, but it has provided the definition now. However it can't provide the specific numerical value for the specific type of parametric resonance. But whatever.

Reply on news.ycombinator.com