I view modern AI as a combination of things that are too early and too late - data science hasn't changed much since the 1990s, but people have a misconception that there is still a lot more to do in data science because of the undeserved excitement around deep learning.