Deep Learning

and also for face detection on digital cameras. CT and MRI scans, and diagnose health conditions. where it is used for localization and mapping, as well as tracking driver state. Perhaps the best-known example of deep learning is DeepMind's AlphaGo. Go is a board game similar to chess. AlphaGo was the first computer program to beat a professional Go player. In March 2016, it beat the top Korean professional Lee Sedol in a match watched by more than 200 million people. The following year, in 2017, AlphaGo beat the world's number one ranking player, China's Ke Jie. In 2016, AlphaGo's success was very surprising. At the time, most people expected that it would take many more years of research before a computer would be able to compete with top-level human Go players. It had been known for a long time that programming a computer to play Go was much more difficult than programming it to play chess. There are many more board configurations possible in Go than there are in chess. This is because Go has a larger board and simpler rules than chess. There are, in fact, more possible board configurations in Go than there are atoms in the universe. This massive search space, and Go's large branching factor the number of board configurations that can be reached in one move makes Go an incredibly challenging game for both humans and computers. One way of illustrating the relative difficulty Go and chess presented to computer programs is through a historical comparison of how Go and chess programs competed with human players. In 1967, MIT's MacHack6 chess program could successfully compete with humans and had an ELO rating well above novice level. And, by May 1997, Deep Blue was capable of beating the chess world champion, Gary Kasparov. In comparison, the first complete Go program wasn't written until 1968 and strong human players were still able to easily beat the best Go programs in 1997. The time lag between the development of chess and Go computer programs reflects the difference in computational difficulty between these two games. However, a second historic comparison between chess and Go illustrates the revolutionary impact that deep learning has had on the ability of computer programs to compete with humans at Go. It took 30 years for chess programs to progress from human-level competence in 1967 to world champion level in 1997. However, with the development of deep learning, it took only 7 years for computer Go programs to progress from advanced amateur to world champion. As recently as 2009, the best Go program in the world was rated at the low end of advanced amateur. This acceleration in performance through the use of deep learning is nothing short of extraordinary. But it is also indicative of the types of progress that deep learning has enabled in a number of fields. AlphaGo uses deep learning to evaluate board configurations and to decide on the next move to make. The fact that AlphaGo used deep learning to decide what move to make next is a clue to understanding why deep learning is useful across so many different domains and applications. Decision making is a crucial part of life. One way to make decisions is to base them on your intuition or your gut feeling. However, most people would agree that the best way to make decisions is to base them on the relevant data. Deep learning enables data-driven decisions by identifying and extracting patterns from large data sets that accurately map from sets of complex inputs to good decision outcomes. Artificial Intelligence, Machine Learning, and Deep Learning Deep learning has emerged from research in artificial intelligence and machine learning. The field of artificial intelligence was born at a workshop at Dartmouth College in the summer of 1956. Research on a number of topics was presented at the workshop, including mathematical theorem proving, natural language processing, planning for games, computer programs that could learn from examples, and neural networks.