Is This Something That Open AI Has Enough GPUs to Do?

This takes a long time like you literally have to retrain the model. It doesn't take into account generality Directly anyway, I think that's a really tough problem. An interesting investigation here would be to play with this in a toy model setting Where actually retraining it like many many times is not as computationally expensive. If we could collect global information about how about people fine-tuning stable diffusion models That's a very similar thing Basically Like every every time someone does a dream booth or something like that if we could actually collect that information and watch the diffusion process Would be a really great research data, but I don't think anyone's recording that right now So yeah, so