Google’s AI solution learns to retain previous knowledge

Google has developed a new algorithm for DeepMind, its artificial intelligence (AI) solution, allowing it to retain knowledge of all of its previously taught skills. As explained by The Merkle, developers were previously able to teach DeepMind to beat one specific game, but once the goal was achieved, they had to wipe its memory to train it for a new game. The Elastic Weight Consolidation (EWC) algorithm allows DeepMind to store information for later use, as well as learn multiple games in succession without problems of delays. The EWC is the first step towards making AI solutions more flexible and efficient, by empowering them with real-time learning capabilities. Solving the ‘conditional forgetting’ problem of AI solutions will allow the AI industry to more forward more rapidly, as systems would be able to retain knowledge about more than one particular field.