In the late '80s I started studying AI in college, fully expecting it to boom. That did not really happen, until it did. Unfortunately, most of what I learned about AI back then is pretty useless now.
I started with symbolic AI: parsers, tree searches, expert systems, logic. Even then it felt like this approach was stuck in a rut. So I changed lanes and studied early ML: decision trees, genetic algo's, nearest neighbours. When I left college SVM's were all the rage.
All that time, neural networks were deemed interesting but unfeasible for anything realistic. Little did we know.
I feel that if I wanted to get back into this field, I might as well start all over again.