[sdubois]

I pretty much agree with what you said, but DL (and even more deep reinforcement learning) is really computationally expensive