General Reinforcement Learning Algorithms

Companies Manipulating A.I. Systems for Competitive Advantage
March 10, 2020
Multitask Learning
March 10, 2020

General Reinforcement Learning Algorithms

Researchers are developing single algorithms that can learn multiple tasks.

Researchers are developing single algorithms that can learn multiple tasks. DeepMind, the team behind AlphaGo, which learned how to play Go with the skill level of a human grandmaster, has developed an innovative new algorithm: AlphaZero.

It is capable of achieving superhuman performance not only in Go, but in other games as well, including chess and shogi (Japanese chess). This one algorithm starts with no knowledge except for the rules of the game and eventually develops its own strategies to beat other players.

In January of 2020, DeepMind published new research showing how reinforcement learning techniques could be used to improve our understanding of mental health and motivation.