In a paper, Google researchers has detailed the way their latest AI evolution, AlphaZero, developed ‘superhuman performance’ in chess, taking only 4 hrs to learn the rules before obliterating the world champion chess program, Stockfish.
We can say, all of humanity’s chess knowledge was absorbed and surpassed by an Artificial Intelligence in almost as long as it takes to drive from New York City to Washington, DC.
After being programmed with the rules of chess (sans strategies), in only 4 hours AlphaZero had mastered the game to the extent it was able to best the highest-rated chess-playing program, Stockfish.
In a series of hundred games against Stockfish, AlphaZero won twenty five games while playing as white (with 1st mover advantage), & picked up 3 games playing as black. The rest of the games were draws, with Stockfish recording no wins & AlphaZero no losses.
Must Read: Samsung gives Honda Civic to winner of prize offer on J-Series Mobile phones
Chess researcher David Kramaley, the CEO of chess science website Chessable says,
“We now know who our new overlord is,” he also adds “It will no doubt revolutionise the game, but think about how this could be applied outside chess. This algorithm could run cities, continents, universes.”
AlphaZero is developed by Google’s DeepMind AI lab tweaked, more generic version of AlphaGo Zero.
For years, DeepMind has been refining this AI, in the process besting a series of human champions who fell just like dominoes before the indomitable, ‘Godlike’ neural network.
That victory streak culminated in a startling success this year in October, in which a fully autonomous new version of the AI – which just learns by playing itself, never by facing humans – bested all of its former incarnations.
AlphaGo Zero’s predecessors did learn partly how to play the game by watching moves been made by human players.
That attempt was intended to help the fledgling AI in learning strategy, but it looks like it may have actually been a handicap, since AlphaGo Zero’s thoroughly self-reliant learning proved devastatingly more than effective in one-on-one competition.
But things move so fast in this field that already the October accomplishment may have been outmoded.
In their recent paper, the team has outlined how the very latest AlphaZero AI takes the self-playing reliance (reinforcement learning) and applies it with a much more generalised streak that gives it broader focus to solve problem.
That broader focus means that AlphaZero does not just play chess. It also plays Shogi and Go too.
For now, Google and computer scientists of DeepMind aren’t commenting publicly on the new research, which has not as yet been peer-reviewed.
