is anyone getting really poor results with chatgpt enrichment in Clay? I'm not having this issue when actually using chatgpt but when i use it in the table, it's not following the instructions properly and completely disregards the examples. For reference, using the same prompts on virgin chats to ensure it's apples to apples
Hey Dave, I'll take a look. I think you're spot on though, sometimes it comes to "prompt engineering" to ensure that the request is as concise as possible and give it a proper structure it understands so it can be much more consistent.
Any specific chatGPT column or all of them are returning bad outputs?
it was the football nickname but it seems to be working now. Very weird. One thing i noticed also was that the prompts was getting half deleted several times as well, in addition to not giving the correct output. All is good now but will keep you posted if that happens again.
Hey Dave, thanks for the follow up. I have not seen that issue before and will try to replicate it on my end. Have you seen this happening in several tables or just this one?
happened in two different types i was working in. I'll send you a screen shot next time it happens. You can see that half of the prompt gets cut off and it followed through to all of the examples.
Thanks, yes, let us know. If we can replicate the behavior then we can figure out a fix for it.