Are the Deep Reinforcement Learning Systems Constrained Enough?

Deep reinforcement learning system could come up with solutions that are just completely unrecognizable in brains. But the problem is, it's very hard to know when you're looking at that. I'm sort of fascinated by these models, because transformers have an arc architectural bias. They have a particular structure that encourages compositional a computation and allows deep learning systems to generalize in ways that they didn't before. Leou just have to ask and find out.