AlphaGo

1、AlphaGODeep-learningsoftwaredefeatshumanprofessionalforfirsttime.WhatisAlphaGOAlphaGoisacomputerprogramdevelopedbyGoogleDeepMindinLondontoplaytheboardgameGo.AlphaGo'salgorithmusesacombinationofneuralnetworks,machinelearningandMonteCarlotreesearchtechniques,combinedwithextensivetraining,bothfromhumanandcomputerplay.WhythegameofGo1017019 × 19150Goisconsideredmuchmoredifficultforcomputerstowinthanothergamessuchaschess,becauseitsmuchlargerbranchingfactormakesitprohibitivelydifficulttousetraditional。

2、AImethods.InChina,JapanandSouthKorea,Goishugelypopularandisevenplayedbycelebrityprofessionals.Theaverage150-movegamecontainsmorepossibleboardconfigurationsthantheatomsintheUniverse.CompetitionsInOctober2015,thedistributedversionofAlphaGodefeatedtheEuropeanGochampionFanHui,a2-danprofessional,fivetozero.ThiswasthefirsttimeacomputerGoprogramhadbeatenaprofessionalhumanplayeronafull-sizedboardwithouthandicap.MatchagainstLeeSedolAlphaGohasdefeatedKoreangrandmasterLeeSedol,finishingthebest-of-fiveserie。

3、swithfourwinsandoneloss.21LeeSedolhadthesecond-highestnumberofGointernationalchampionshipvictoriesintheworld.Butafterthiscompetition,AlphaGoreplaceshisranking.ResponsestovictorymilestoneprogressAlphaGo'sMarch2016victorywasamajormilestoneinartificialintelligenceresearch.Gohadpreviouslybeenregardedasahardprobleminmachinelearningthatwasexpectedtobeoutofreachforthetechnologyofthetime.WhencomparedwithDeepBlue,AlphaGo'sunderlyingalgorithmsarepotentiallymoregeneral-purpose,andmaybeevidencethatthescient。

4、ificcommunityismakingprogresstowardartificialgeneralintelligence.。