> 63. When we write programs that "learn", it turns out that we do and they don't. This was what turned me off of machine learning some 30 years after this was

> 63. When we write programs that "learn", it turns out that we do and they don't. This was what turned me off of machine learning some 30 years after this was written. Deep learning is amazing, but it still seems to be the people tweaking the experimental setup who are doing the real learning.