I’ve been prompting ChatGPT (“gpt-3.5-turbo”) with the exact same prompts all day and suddenly, in the past hour or so, the responses became very different (in a bad way). ChatGPT started completely ignoring response format constraints, returning empty output, and unusual data when earlier in the day it had been giving good responses.
Has anyone else noticed this? I’m hesitant to change any of my prompts until I try again tomorrow in case this is a bug that OpenAI is working on.