资源描述：
《Neuron-2017-Neuroscience-Inspired Artificial 》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、NeuronReviewNeuroscience-InspiredArtificialIntelligenceDemisHassabis,1,2,*DharshanKumaran,1,3ChristopherSummerfield,1,4andMatthewBotvinick1,21DeepMind,5NewStreetSquare,London,UK2GatsbyComputationalNeuroscienceUnit,25HowlandStreet,London,UK3InstituteofCogni
2、tiveNeuroscience,UniversityCollegeLondon,17QueenSquare,London,UK4DepartmentofExperimentalPsychology,UniversityofOxford,Oxford,UK*Correspondence:dhcontact@google.comhttp://dx.doi.org/10.1016/j.neuron.2017.06.011Thefieldsofneuroscienceandartificialintelligen
3、ce(AI)havealongandintertwinedhistory.Inmorerecenttimes,however,communicationandcollaborationbetweenthetwofieldshasbecomelesscommonplace.Inthisarticle,wearguethatbetterunderstandingbiologicalbrainscouldplayavitalroleinbuildingintelligentmachines.Wesurveyhi
4、storicalinteractionsbetweentheAIandneurosciencefieldsandemphasizecurrentadvancesinAIthathavebeeninspiredbythestudyofneuralcomputationinhumansandotheranimals.Weconcludebyhighlightingsharedthemesthatmaybekeyforadvancingfutureresearchinbothfields.Inrecentyear
5、s,rapidprogresshasbeenmadeintherelatedtively.Forexample,ifanalgorithmisnotquiteattainingthelevelfieldsofneuroscienceandartificialintelligence(AI).Attheofperformancerequiredorexpected,butweobserveitiscoretodawnofthecomputerage,workonAIwasinextricablyinter-t
6、hefunctioningofthebrain,thenwecansurmisethatredoubledtwinedwithneuroscienceandpsychology,andmanyoftheearlyengineeringeffortsgearedtomakingitworkinartificialsystemspioneersstraddledbothfields,withcollaborationsbetweenarelikelytopayoff.thesedisciplinesprovin
7、ghighlyproductive(ChurchlandandOfcoursefromapracticalstandpointofbuildinganAISejnowski,1988;Hebb,1949;Hintonetal.,1986;Hopfield,system,weneednotslavishlyenforceadherencetobiological1982;McCullochandPitts,1943;Turing,1950).However,plausibility.Fromanengine
8、eringperspective,whatworksismorerecently,theinteractionhasbecomemuchlesscommon-ultimatelyallthatmatters.Forourpurposesthen,biologicalplace,asbothsubjectshavegrownenormouslyincomplexityplausibilityisaguide,notastrictrequire
