I’ve been using ChatGPT 4 for a while, and I have a a library of customized prompt templates that work perfectly. However, today, a prompt I’ve used very successfully in the past, suddenly returned a nonsense response. The prompt was in two parts, ChatGPT requested the second prompt after I input the first, but instead of completing the prompt in the template I was using, it went off the rails and did something completely different. I haven’t been able to get it to work properly all afternoon. Does anybody have any advice? I use ChatGPT professionally, and I don’t want my productivity to suffer. (Or my sanity, for that matter). Thanks ahead of time!
1 Like
While I don’t have a generic solution
I can confirm that a lot of people confirmed prompt drift woth GPT-4 recently
Thank you. I tried the prompt again today and it worked just fine. I’m hoping this doesn’t become a recurring issue.
A ton of people in my circle are complaining about GTP-4 getting “dumber” or at least significantly worse than back a few months as well, and they heading towards the supposedly much better Bard. One theory is because Open AI are cutting costs by running a different much cheaper model than actually set. Stanford ran a study comparing the two
1 Like