Rebooting AI: Building Artificial Intelligence We Can Trust

Twoleadersinthefieldofferacompellinganalysisofthecurrentstateoftheartandrevealthestepswemusttaketoachieveatrulyrobustartificialintelligence.DespitethehypesurroundingAI,creatinganintelligencethatrivalsorexceedshumanlevelsisfarmorecomplicatedthanwehavebeenledtobelieve.ProfessorsGaryMarcusandErnestDavishavespenttheircareersattheforefrontofAIresearchandhavewitnessedsomeofthegreatestmilestonesinthefield,buttheyarguethatacomputerbeatingahumaninJeopardy!doesnotsignalthatweareonthedoorstepoffullyautonomouscarsorsuperintelligentmachines.Theachievementsinthefieldthusfarhaveoccurredinclosedsystemswithfixedsetsofrules,andtheseapproachesaretoonarrowtoachievegenuineintelligence.Therealworld,incontrast,iswildlycomplexandopen-ended.Howcanwebridgethisgap?Whatwilltheconsequencesbewhenwedo?Takinginspirationfromthehumanmind,MarcusandDavisexplainwhatweneedtoadvanceAItothenextlevel,andsuggestthatifwearewisealongtheway,wewon'tneedtoworryaboutafutureofmachineoverlords.Ifwefocusonendowingmachineswithcommonsenseanddeepunderstanding,ratherthansimplyfocusingonstatisticalanalysisandgatherineeverlargercollectionsofdata,wewillbeabletocreateanAIwecantrust—inourhomes,ourcars,andourdoctors'offices.RebootingAIprovidesalucid,clear-eyedassessmentofthecurrentscienceandoffersaninspiringvisionofhowanewgenerationofAIcanmakeourlivesbetter.