Its getting down right Orwellian out there...

A new chess whiz
Dec10 2018 shogi
Deep Blue became the first computer to beat a chess grandmaster in 1997. In the 20+ years since, artificial intelligence programmers have been looking for the next Holy Grail: an algorithm that can teach itself how to play chess. It looks like they just found it. Announced last week in Science, the company DeepMind has developed a program called AlphaZero that can beat the world's best chess players. It can also beat Stockfish, the leading chess-playing program. It won 155 out of 1,000 games, losing only six—the rest were draws. But what's remarkable is that AlphaZero has also learned how to play Shogi, a Japanese cousin of chess. And the Chinese game Go, which has 300 times more possible moves than chess. Unlike previous game-playing algorithms, AlphaZero doesn't start with a huge dataset of human knowledge. Instead, it just starts with a model of the game rules and begins playing games against a computer. As it plays thousands of games, it builds its own algorithm based on trial and error. To do this, AlphaZero uses 5,000 computers, or TPUs, developed by DeepMind's parent company Google specifically for AI programs. Using that processing power, AlphaZero was able to play 60 million games of chess in just a few hours and then beat Stockfish. It beat the reigning computer champions in Go and Shogi with similar amounts of effort. Next up, AlphaZero hopes to master Texas hold 'em and multiplayer video games. These are much trickier problems since not all information is available to all players. The scientists hope their effort isn't just fun and games. They hope their work could be directly applicable to medicine, math, and scientific research. Not to mention it's already teaching chess grandmaster lots of new moves.
More about this Curio:
Scientific American: ""Superhuman" AI Triumphs Playing the Toughest Board Games"
Ars Technica: "Move over AlphaGo: AlphaZero taught itself to play three different games"
Science: "Chess, a Drosophila of reasoning"
Science: "A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play"
PRI: "Garry Kasparov and the game of artificial intelligence"