strong lotus
Hey, it's a bit hard for me to prove but i've noticed, at least today that GPT-4 is giving responses a lot faster than usual. Whilst I was initially delighted, the more I realized the responses were the same level of briefness and inaccuracy as 3.5.
This was compounded by the fact that it only seemed to comprehend the first and second sentence of the prompt I was giving, if I put these sentences further down the prompt and told it to do other things first (Like in a list) it'll completely ignore them and not do it.
I'm not sure if this is a mistake or if this is your method of stopping the servers blowing up. If this IS intentional then it's very scammy considering people pay for GPT-4 access.
hoary echo