Take the example of the AlphaGo system, which was developed by UK-based start-up DeepMind, which leverages an acute flavor of deep reinforcement learning to inform its predictions. There is good reason behind Google's move to acquire it for a round sum of $500 million, since many claim that DeepMind has made first steps toward something called Artificial General Intelligence (AGI)—sort of the Holy Grail of AI, if you will. This notion refers to the capability of an artificially intelligent system to perform well on various tasks, instead of the narrow span of application our networks have taken so far. A system that learns through observing its own actions on an environment is similar in spirit (and potentially much faster) to how we humans learn ourselves.
The networks we built in the previous chapters perform well at a narrow...
