Not every problem can be solved by matrix multiplication and gradient descent. Particularly, many problems require exact solutions, and not just good approximat

Not every problem can be solved by matrix multiplication and gradient descent. Particularly, many problems require exact solutions, and not just good approximations (think aerospace or science or finance). And speaking of science particularly, many solutions require insight, which ML often can't provide.