>>712556548
this is what's necessary for the training process, at least right now
openai has to do this too with people in africa looking through prompts, it's called rlhf (reinforcement learning through human feedback), some of them have even said it's given them ptsd looking through some of the shit
it doesn't look like it will always need to be that way though, the reinforcement learning appears to be applied through humans willingly offering feedback and continued use and training data when they use the models, without even knowing that's what they're doing