This is nothing new, tho. The downfalls of reward maximization has been a known issue without a solution ever since reinforcement learning was first researched.. in the 1980s.