Heighn comments on Reward is not the optimization target

[Heighn]

“And not only do I not expect the trained agents to not maximize the original “outer” reward signal”

Nitpick: one “not” too many?

[TurnTrout]

Thanks, fixed.