> You can show a person a couple pictures of a dog, and they can construct a 3-D mental model of a dog, and what it looks like from all angles and in all poses.

> You can show a person a couple pictures of a dog, and they can construct a 3-D mental model of a dog, and what it looks like from all angles and in all poses. Yet you have to give thousands of examples to do the same thing for a ML/DL system. By the time a human brain can do this, aren't they also exposed to thousands of examples? How do we learn to speak for example? It's a long painful process of trial and error. How is this significantly different from a supervised learning algorithm? Teaching a child to catch is very similar. At first, they can't recognize the trajectories of thrown objects very well and they aren't coordinated enough to move their hands to where the projectile is. After time and many repetitions, they get to the point where they can catch a ball without consciously following the ball or can even catch without looking based on initial trajectory.