DeepMind’s MuZero ups the reinforcement learning ante

In “a significant step forward in the pursuit of general-purpose algorithms” that “are able to deal with the messiness and complexity of the real world”, Google DeepMind‘s new reinforcement learning algorithm, MuZero, which like its predecessor AlphaGo uses lookahead search, can achieve superhuman levels of prowess at Go, chess and video games without any prior knowledge of the rules. Details were published in the scientific journal Nature just before Christmas.

Source: The Independent