[djeastm]

I thought reinforcement learning with human feedback was meant to get that quantification of "taste"