It seems like ChatGPT with the GPT-4 model is much worse now? It seems to ignore, or not pay “attention” to parts of my prompts, and is worse at following instructions than before. Anybody else noticing this, or am I just imagining it?
Last week, my credit card company misclassified my subscription as fraud, which caused my subscription lapse (I started my subscription back in January, I believe). Seems like resubscribing coincided with me noticing worse performance.
Edit: I mostly use ChatGPT to produce personalized code examples, or do code “translation” and refactoring.
they pushed a new fine tune a month back, people arent really liking it
Yeah, it’s gotten much worse. So many hallucinations now that it might as well be 3.5.
I stick with 3.5 just because it’s faster. I tried 4 a few times, but i didn’t feel it justified the slowness with quality.
On a positive note. I am using the API with 3.5-turbo in an app prototype, and it seems to be following instructions better than it used to. For ChatGPT, I don’t really care to sacrifice speed for quality in my use-cases.
3.5 recently had an update that did that, before that it was really bad.
