您好，欢迎访问三一刀客
当前位置：首页 > 研究报告 > 浅谈人工智能下一个十年-清华大学-202004
1浅谈人工智能的下个十年JieTangComputerScienceTsinghuaUniversity2人工智能的第三次浪潮3人工智能历史4人工智能领域发展趋势PoweredbyClaudeShannonShannon,ClaudeE.XXII.Programmingacomputerforplayingchess.Philosophicalmagazine41.314(1950):256-275.AlanTuringTuring,AlanM.Solvableandunsolvableproblems.ScienceNews-ens.fr39(1954).1950计算机象棋博弈1954图灵测试5人工智能领域发展趋势PoweredbyJohnMcCarthyMcCarthy,J.,etal.DartmouthConference.DartmouthSummerResearchConferenceonArtificialIntelligence.19561956达特茅斯会议1959一般问题解决器MarvinMinskyNathanRochesterClaudeShannonHerbertSimonJ.C.ShawAllenNewellNewell,A.;Shaw,J.C.;Simon,H.A.(1959).Reportonageneralproblem-solvingprogram.ProceedingsoftheInternationalConferenceonInformationProcessing.pp.256–264.6人工智能领域发展趋势PoweredbyDanielBobrowBobrow,DanielG.Naturallanguageinputforacomputerproblemsolvingsystem.(1964)JosephWeizenbaumWeizenbaum,Joseph.ELIZA—acomputerprogramforthestudyofnaturallanguagecommunicationbetweenmanandmachine.CommunicationsoftheACM9.1(1966):36-45.1964理解自然语言输入1966ELIZA人机对话7人工智能领域发展趋势PoweredbyBuchanan,Bruce,GeorgiaSutherland,andEdwardA.Feigenbaum.HeuristicDENDRAL:aprogramforgeneratingexplanatoryhypothesesinorganicchemistry.DefenseTechnicalInformationCenter,1968.1968世界首个专家系统DENDRALEdwardFeigenbaum8人工智能领域发展趋势PoweredbyRandallDavisApplicationsofmetalevelknowledgetotheconstruction,maintenanceanduseoflargeknowledgebases[M].StanfordUniversity,ComputerScienceDepartment,AILaboratory,1976.DrewMcDermott,JonDoyleMcDermottD,DoyleJ.Non-monotoniclogicI[J].Artificialintelligence,1980,13(1):41-72.1976大规模知识库构建与维护1980非单调逻辑9人工智能领域发展趋势PoweredbyBerlinerHJ.Backgammoncomputerprogrambeatsworldchampion[J].ArtificialIntelligence,1980,14(2):205-220.1980计算机战胜双陆棋世界冠军HansBerliner10人工智能领域发展趋势PoweredbyBrooksR.Arobustlayeredcontrolsystemforamobilerobot[J].RoboticsandAutomation,IEEEJournalof,1986,2(1):14-231987基于行为的机器人学RodneyBrooks11人工智能领域发展趋势PoweredbyTesauroG.TD-Gammon,aself-teachingbackgammonprogram,achievesmaster-levelplay[J].Neuralcomputation,1994,6(2):215-219.1987自我学习双陆棋程序GerryTesauro12人工智能领域发展趋势PoweredbyTimBerners-LeeBerners-Lee,Tim.Semanticwebroadmap.(1998).McGuinness,DeborahL.,andFrankVanHarmelen.OWLwebontologylanguageoverview.W3Crecommendation10.2004-03(2004):10.1998语义互联网路线图2004OWL语言13人工智能领域发展趋势PoweredbyGeoffreyHintonHinton,GeoffreyE.,SimonOsindero,andYee-WhyeTeh.Afastlearningalgorithmfordeepbeliefnets.Neuralcomputation18.7(2006):1527-1554.Le,QuocV.,etal.Buildinghigh-levelfeaturesusinglargescaleunsupervisedlearning.arXivpreprintarXiv:1112.6209(2011).2006深度学习2011高层抽象特征构建14人工智能领域发展趋势PoweredbyMarkoff,John.Googlecarsdrivethemselves,intraffic.TheNewYorkTimes10(2010):A1.2009谷歌自动驾驶汽车SebastianThrun15人工智能领域发展趋势PoweredbyIBM’sWatsonMarkoff,John.Computerprogramtotakeon‘Jeopardy!’.TheNewYorkTimes(2009).Apple’sSiriSadun,Erica,andSteveSande.TalkingtoSiri:LearningtheLanguageofApple'sIntelligentAssistant.QuePublishing,2013.2011沃森获得Jeopardy冠军2011自然语言问答16人工智能领域发展趋势Poweredby17人工智能近10年18AI趋势：从感知到认知•FromperceptrontocognitionComputingPerceptionCognitionStorage&ComputingRecognizetext,images,objects,voicesOrganizeandgenerateknowledge,reasoning19ArtificialIntelligenceAlphaGoImagerecognitionSelf-driving20DDPG(2015)A3C(2016)Perceptron(1958)FrankRosenblattCornellUniversitypsychologistBPNN/MLP(1986)HopfieldNetwork(1982)[recurrent&feedback]GeofferyHintonUniversityofTorontodeeplearningNeocognitron(1980)[convolution&pooling]LeNet/CNN(1998)YannLecunNewYorkUniversitydeeplearningAlexNet(2012)Relu,dropout&biggerVGG(2014)GoogLeNet(2015)ResNet(2016)KaimingHeMSRA=FAIRcomputervisionDenseNet(2017)RBM(1986/2006)DeepBeliefNets(2006)stackAutoEncoder(1989/2006)DenosingAutoencoder(2008)VAE(2013)VariationalInferenceMaxWellingUniversityofAmsterdamstatisticallearningGAN(2014)DCGAN(2014)WGAN(2017)PGGAN(2017)IanGoodfellowGoogleBraindeepadversariallearningRNN/LSTM(1997)JürgenSchmidhuberIDSIAUniversalAISeq2Seq(2014)RNNinSpeechRecognition(2013)YoshuaBengioUniversityofMontrealDeeplearningNeuralProbabilisticLanguageModel(2003)word2Vec(2013)SeqGAN(2017)LeakGAN(2018)CharacterCNN(2015)self-attention(2017)DeepQ-learning(2013)AlphaGo(2016)DoubleDQN(2015)DuelingNet(2016)DavidSilverDeepMindReinforcementlearningAlphaZero(2017)CapsuleNets(2017)算法21BERT•Pre-train•Finetune•Beatallstate-of-the-artson11NLPtasksin2018•AutoregressiveModel•BeatBERTin2019•ALiteBERT•Parameter-reductiontechniques•BeatXLNetandalltheothers•Thebestvideosynthesisperformance•ByDeepMind•Unsupervisedvisualrepresentationlearning•Momentumcontrastivelearning•Outperformitssupervisedpre-trainingcounterparts•Simplifiedcontrastivelearningframework•Outperformpreviousself-supervisedandsemi-supervisedmethodsonImageNet人工智能未来…29第三代人工智能的理论体系•早在2015年，张钹老师就提出第三代人工智能体系的雏形；2017年DARPA发起XAI项目，从可解释的机器学习系统、人机交互技术以及可解释的心理学理论三个方面，全面开展可解释性AI系统的研究•2018年底，正式公开提出第三代人工智能的理论框架体系–建立可解释、鲁棒
三一刀客所有资源均是用户自行上传分享，仅供网友学习交流，未经上传用户书面授权，请勿作他用。
37页
79页
50页
41页
18页
32页
39页
55页
43页
80页
扫描二维码
pei**com
本文标题：浅谈人工智能下一个十年-清华大学-202004
链接地址：https://www.111doc.com/doc-4936371 .html