Further two things.
By deep learning, I'm talking about the method of AI which is famous for learning and playing Atari video games by being given some screenshots of desirable end results. I think most AI that's up and running now, which is proving to be of any use, harks back to, is based on, deep learning?
And just to actually state it clearly, my main question is: Could deep learning solve Winograd schema problems or not do you think?