This was a fairly simple problem that even GPT-3.5 was able to solve in April. Now, while the provided solution seems accurate at first sight, it seems to go into an infinite answer loop in the middle of issuing me a caveat: https://chat.openai.com/share/4af022b9-c58e-4ac3-87f7-a4e56cd8ea6c<p>"...can obtain these from the Google Developer Console by setting up a new OAuth 2.0 Client ID. Be sureSure, I'll outline a basic web application setup..."<p>Is anyone else experiencing this sort of issue lately?
Interesting. "Looping" is a common issue with LLaMA too, and in my experience changing certain generation parameters (without any changes to the model) make it worse. Maybe OpenAI adjusted something on their end?