Comment by ethbr1

Isn't that just RL with extra power-intensive steps? (An entire model chugging away in the goal function)