> why did it take us so long to develop advanced neural networks?
I have a pet theory (with some empirical support) that Marvin Minsky put somewhat of a multi-year kibosh on "perceptron"/neural-net research due to his (true, but arguably irrelevant) assertion that trained neural nets were essentially "nondeterministic black boxes" which was basically heretical.
is a brilliant book from the early days of computer science which is asking questions about the fundamentals of algorithms and what you can and can't do with different methods. (Like the result that it takes at least N log N comparisons to sort N items)
I have a pet theory (with some empirical support) that Marvin Minsky put somewhat of a multi-year kibosh on "perceptron"/neural-net research due to his (true, but arguably irrelevant) assertion that trained neural nets were essentially "nondeterministic black boxes" which was basically heretical.