In order to improve the learning rate in robots, Google’s AI company DeepMind is pioneering a new technology which allows robots to dream. The substance of these AI dreams consists primarily of scenes from Atari Video games. Google’s first major successes with DeepMind involves teaching AI to play ancient videos games like Breakout and Asteroids. But the end game here is for robots to dream about much the same things humans do – experiment and analyze how different courses of action affect the outcome.
When robots can already take over humans in most games such as Chess and Go, you might ask why AI “dreams” are necessary. To understand this, it is essential to distinguish between AIs that use supervised vs. unsupervised learning. Most of the remarkable achievement so far attained by AI have been made using supervised learning, in which prearranged “training data” is supplied by the programmers and the AI learns to detect patterns within the data. This is a quite clear-cut approach to teaching machines but definitely not how humans learn. We use an approach more similar to unsupervised learning in which agent experiments on their own to find out how different courses of action influence their goals. As it involves experimentation, This type of learning is far more time consuming than supervised learning but the folks at DeepMind are mainly concerned with unsupervised learning because they believe that it holds the best hope for creating AI with human-like general intelligence.