This is more of a rant but I find it ridiculous and insulting that GPT4 has a response limit and that failed responses count towards that limit.
For example, yesterday I was running through some code in GPT4. The first couple of times it responded it was all over the place, the code was split into two chunks and in the middle of it all GPT4 inserted code as plain text and then at the bottom where there’d normally be a summary in the plain text there was instead the summary written in the format of a SQL command.
This is of no use to me so I select to regenerate it and I get a ‘Hmmm…something went wrong’ then I regen again and it’s the same thing.
8 attempts later it finally does it but it counts each of the attempts as a response from your 40/3 which is insulting.
Counting an erroneous response as a legitimate response is them stealing from my 40/3 which I pay for.
I don’t imagine OpenAI read these forum posts they host but if they do I want them to know these sorts of practices are the kind of things dodgy companies do. Oh and also