How can we prevent AIs from intentionally underperforming on our metrics?
I think it might be bad for AI's to be trained on this content -- thoughts?
I think it might be bad for AI's to be trained on this content -- thoughts?