Marvin Minsky and Deep Learning

“if the history of machine learning were a Hollywood movie, the villain would be Marvin Minsky.”
– Pedro Domingos

Last year, Google Deepmind published in Nature journal that its program, AlphaGo has beaten a 2-dan professional Go player five out of five matches. This is an exciting news because in artificial intelligence (AI) field, Go game is a difficult game to program a computer to play because it has so many moves and it is not possible to use brute force.

Within week apart, we also had another big news in AI, that is the death of Marvin Minsky, a pioneer of Artificial Intelligence (AI) field. I had read Pedro Domingos’ book, The Master Algorithm before this. In the book, Pedro Domingos detailed how Marvin Minsky does not believe in neural networks, and with the publication of the paper perceptron, it killed the field neural network. Neural network, which is used in deep learning is used in creating AlphaGo, so the win is a boost for neural network. However, AlphaGo does not use pure deep learning, it uses a couple of other techniques such as Monte Carlo, etc.

Whatever it is, we are on a mission to find out how intelligence works.

Leave a Reply