Using a bio-inspired system architecture, Deep Mind scientists have created a single algorithm (Deep Q-network: “deep convolutional network”) that is actually able to develop problem-solving skills and is able to understand spatial relationships between different objects in an image, such as distance from one another, in such a sophisticated way that it can actually re-envision the scene from a different viewpoint. This type of system was inspired by early work done on the visual cortex. And then they immediately put it to use learning a set of classic Atari video games.
- Nature paper: https://www.nature.com/articles/nature14236
- Ars Technica article: https://arstechnica.com/science/2015/02/ai-masters-49-atari-2600-games-without-instructions/
- only visual images and scores as input
- same algorithm on each of 49 games
- 75% of human score on majority of games
- some generalizing from previous experiences