1、 超级计算学院 人工智能 Chapter01 AboutAI 人工智能全册配套课件人工智能全册配套课件 超级计算学院 人工智能 Chapter01 AboutAI Introduction toArtificial Intelligence 超级计算学院 人工智能 Chapter01 AboutAI 人工智能及其应用人工智能及其应用 Artificial Intelligence and Its Application 超级计算学院 人工智能 Chapter01 AboutAI 任课教师: 姜云飞 1978-1997 吉林大学计算机科学系硕士研究生, 讲师, 副教授, 教授, 计算机科学系副主
2、任, 计算机研究 所副所长, 博士导师, 期间曾到上海外国语大学, 英国Ulster大学, 美国新泽西理工学院(NJIT)学 习。 1997- 中山大学信息科学与技术学院教授, 博士导师 1999-2003 信息科学与技术学院院长 2004-2007 广东省计算机学会理事长主要研究领域: 人工智能, 知识工程, 智能诊断, 智能规划 超级计算学院 人工智能 Chapter01 AboutAI How to learn the course AI? 1. You should be active. 2. You should do meditation. 3. You should read s
3、omething by yourself. 4. You should take lessons 5. You should do some exercises 超级计算学院 人工智能 Chapter01 AboutAI Text book:Artificial Intelligence: Structures and Strategies for Complex Problem Solving(English version 5)George F. Luger 超级计算学院 人工智能 Chapter01 AboutAI 1。 什么是人工智能?什么是人工智能? What is Artifici
4、al Intelligence?What is Artificial Intelligence? 2 2。 人工智能的发展历史人工智能的发展历史 The history of Artificial Intelligence.The history of Artificial Intelligence. 3 3。 人工智能的研究领域。人工智能的研究领域。 The fields of Artificial Intelligence.The fields of Artificial Intelligence. 4 4。 人工智能的应用人工智能的应用 The Applications of Artif
5、icial IntelligenceThe Applications of Artificial Intelligence8吉林大学珠海学院计算机科学与技术系人工智能 5 5。 人工智能在中国人工智能在中国 Artificial Intelligence in ChinaArtificial Intelligence in China6 6。 我们的研究工作我们的研究工作 Our Research works in AIOur Research works in AI7 7。 人工智能对人类的挑战人工智能对人类的挑战 Challenge of AI to MankindChallenge of
6、 AI to Mankind 超级计算学院 人工智能 Chapter01 AboutAI 1. . 什么是人工智能?什么是人工智能? What is Artificial Intelligence?What is Artificial Intelligence?1.1 什么是智能什么是智能 ? What is Intelligence? Intelligence is the computational part of the ability to achieve goals in the world. Varying kinds and degrees of intelligence occ
7、ur in people, many animals and some machines. McCarthy 超级计算学院 人工智能 Chapter01 AboutAI 人类的主要智能行为:The main intelligent behavioursComputation, analysis, reasoning, synthesis, decision, Playing games, learning, talking.Among these mind actions, the reasoning plays key roles.Reasoning is more complex than
8、 other actions 智能: 人类在处理复杂问题时表现出的思维能力。 计算, 分析, 推理, 综合, 决策, 诊断, 游戏, 学习,谈话 推理是一种其核心作用的智能行为. 推理比计算更复杂. 超级计算学院 人工智能 Chapter01 AboutAI Intelligence: The minding ability manifested by mankind when he deals with the complicated events. 超级计算学院 人工智能 Chapter01 AboutAI 1.2 What is Artificial Intelligence ?It i
9、s the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable. McCarthy(12 November 2007
10、). What Is Artificial Intelligence? 超级计算学院 人工智能 Chapter01 AboutAI 人工智能即包含理论研究的内容又包含工程方面的内容.人工智能的研究注意智能系统的效果而不是单纯的对人的智能行为的模拟. 人们看到鸟在天空中飞翔, 想到要造能飞行的机器,但现在的飞机使用的空气动力学原理与鸟的飞行道理有很大的不同,同样实现了在天上飞的梦想,而且飞的更快,更高,更远. 人工智能并不是一个只是在高楼深院里才能研究的人工智能并不是一个只是在高楼深院里才能研究的科学。科学。 它能使计算机系统的性能得到明显提高它能使计算机系统的性能得到明显提高。人人工智能的应用
11、将把人类带进一个崭新的时代。工智能的应用将把人类带进一个崭新的时代。 超级计算学院 人工智能 Chapter01 AboutAI 1.3 Turing TestAlan Turings 1950 article Computing Machinery and Intelligence Tur50 discussed conditions for considering a machine to be intelligent. He argued that if the machine could successfully pretend to be human to a knowledgeab
12、le observer then you certainly should consider it intelligent. 超级计算学院 人工智能 Chapter01 AboutAI The Turing test(http:/www.abelard.org/turpap/turpap.htm)Alan Turing (1950) “Computing Machinery and Intelligence”a machine may be deemed intelligent, if it can act in such a manner that a human cannot distin
13、guish the machine from another human merely by asking questions via a mechanical link. 超级计算学院 人工智能 Chapter01 AboutAI ? ?MachineMachine 超级计算学院 人工智能 Chapter01 AboutAI 2。 人工智能的发展历史人工智能的发展历史 The history of Artificial Intelligence.AI is nearly as old as computing. 超级计算学院 人工智能 Chapter01 AboutAI The Gestat
14、ion of Artificial Intelligence(1943-1956) 人工智能发展的初期人工智能发展的初期Warran McCulloch and Walter Pitts Model of Artificial NeuronsAlan Turing chess program(1953)Marvin Minskey and Dean Edmond (SNARC, 1951) The first neural network computer 3000 vacuums 超级计算学院 人工智能 Chapter01 AboutAI Minskeys Ph.D. committee w
15、as skeptical whether this kind of work should be considered mathematicsbut von Neumann:”If it isnt now, it will be some day” 超级计算学院 人工智能 Chapter01 AboutAI Dardmouth conferencrconferencr(1956)(1956)John McCarthyMinskey, Shannon, Samuel(IBM), Solomonof, Selfridge(MIT) 10 attendeesAlen Newell and Herbe
16、rt Simon(Carnegie Tech)Logic Theoristprove most of the theorems in Chapter 2 of Russells principia Mathematica Rejected by the editor of Journal of Symbolic logic 超级计算学院 人工智能 Chapter01 AboutAI John McCarthy (September 4, 1927 October 24, 2011) 超级计算学院 人工智能 Chapter01 AboutAI John McCarthy (September 4
17、, 1927 October 24, 2011) was an American computer scientist and cognitive scientist. McCarthy was one of the founders of the discipline of artificial intelligence. He coined the term artificial intelligence (AI), developed the Lisp programming language family, significantly influenced the design of
18、the ALGOL programming language, popularized timesharing, and was very influential in the early development of AI.McCarthy received many accolades and honors, such as the Turing Award for his contributions to the topic of AI, the United States National Medal of Science, and the Kyoto Prize. 超级计算学院 人工
19、智能 Chapter01 AboutAI Herbert Alexander Simon (June 15, 1916 February 9, 2001), a Nobel laureate, was an American political scientist, economist, sociologist, psychologist, computer scientist, and Richard King Mellon Professormost notably at Carnegie Mellon Universitywhose research ranged across the
20、fields of cognitive psychology, cognitive science, computer science, public administration, economics, management, philosophy of science, sociology, and political science, unified by studies of decision-making. With almost a thousand highly cited publications, he was one of the most influential soci
21、al scientists of the twentieth century. 超级计算学院 人工智能 Chapter01 AboutAI Simon was among the founding fathers of several of todays important scientific domains, including artificial intelligence, information processing, decision-making, problem-solving, attention economics, organization theory, complex
22、 systems, and computer simulation of scientific discovery. 超级计算学院 人工智能 Chapter01 AboutAI Early enthusiasm, great expectation(1952-1969)早期的狂热和大胆的预测早期的狂热和大胆的预测The early years of AI were full of success!Newell and Simons GPS(Genaral problem Solver)Arthur Samuels Checker(TV,Feb. 1956)Slagle integration
23、system SAINT, undergraduates levelINT experts level 超级计算学院 人工智能 Chapter01 AboutAI Computer checkers (draughts) developmentSamuel is most known within the AI community for his groundbreaking work in computer checkers in 1959, and seminal research on machine learning, beginning in 1949. He thought tha
24、t teaching computers to play games was very fruitful for developing tactics appropriate to general problems, and he chose checkers because it is relatively simple, but has a depth of strategy. The main driver of the machine was a search tree of the board positions reachable from the current state. S
25、ince he had only a very limited amount of available computer memory, Samuel implemented what is now alpha-beta pruning. Instead of searching each path until it came to the games conclusion, 超级计算学院 人工智能 Chapter01 AboutAI Samuel developed a scoring function based on the position of the board at any gi
26、ven time. This function tried to measure the chance of winning for each side at the given position. It took into account such things as the number of pieces on each side, the number of kings, and the proximity of pieces to being “kinged”. The program chose its move based on a minimimax strategy, mea
27、ning it made the move that optimized the value of this function, assuming that the opponent was trying to optimize the value of the same function from its point of view. 超级计算学院 人工智能 Chapter01 AboutAI Samuel also designed various mechanisms by which his program could become better. In what he called
28、rote learning, the program remembered every position it had already seen, along with the terminal value of the reward function. This technique effectively extended the search depth at each of these positions. Samuels later programs reevaluated the reward function based on input from professional gam
29、es. He also had it play thousands of games against itself as another way of learning. With all of this work, Samuels program reached a respectable amateur status, and was the first to play any board game at this high a level. He continued to work on checkers until the mid-1970s, at which point his p
30、rogram achieved sufficient skill to challenge a respectable amateur. 超级计算学院 人工智能 Chapter01 AboutAI Robinson(1956) resolution principle machine translation: 80% A series of victories lead to Optimistic estimate It is not my aim to surprise or shock you- but simplest way I can summarize is to say that
31、 there are now in the world machine that think, that learn and that create. Herbert Simon, 1957 超级计算学院 人工智能 Chapter01 AboutAI A dose of reality(1966-1972)现实的药方现实的药方 no progress of Samuels Checker Machine translation: 20% difficult an wrong example of machine translation The spirit is willing but the
32、 fresh is weak The vodka is willing but the fresh is rotten The cause of failure short of understand, search without guide, Knowledge 超级计算学院 人工智能 Chapter01 AboutAI Knowledge-based system: The key to power?1969-1979Knowledge is power.FeigenbaumDENDRAL , MYCIN, ProspectRutgers University Amarelproject
33、 computer in biomedicineRoger Schank at Yale, “There is no such thing as syntax, ” “all needed are understanding.” 超级计算学院 人工智能 Chapter01 AboutAI AI becomes an industry(1980-1988)The first successful commercial expert system:R1,DEC, McDermott, $40 million a year(1986)1988, 100 in use, 500 in developm
34、ent from a few million in 1980 to $2 billion in 19881997, IBMs blue defeated the Kaspalov in Chess 超级计算学院 人工智能 Chapter01 AboutAI 新浪科技新浪科技 2016.02. 27谷歌人工智能挑战人类围棋冠军详情:将比赛谷歌人工智能挑战人类围棋冠军详情:将比赛5场场新浪科技讯 北京时间2月22日消息,继宣布AlphaGo实现突破性研究计算机程序首次击败专业棋手之后, Google DeepMind今日公布了即将与过去十年最佳围棋手李世石之间的终极挑战的详细情况。3月9日至3月1
35、5日,AlphaGo将在韩国首尔与李世石进行5场挑战赛。比赛完全平等,获胜者将得到一百万美元奖金。如果AlphaGo获胜,奖金将捐赠给联合国儿童基金会(UNICEF),STEM教育,以及围棋慈善机构(Go Charity)。 超级计算学院 人工智能 Chapter01 AboutAI 因围棋步骤的绝对数量比宇宙的原子数还多,它一直被视为最复杂的电脑游戏之一,也是人工智能始终未解的挑战。DeepMind在上月的科学杂志Nature,以一篇论文公布了这一突破性进展的详细情况。比赛将于北京时间中午12点在首尔四季酒店举行,具体日程如下:1. 3月9日 (星期三):首场比赛2. 3月10日(星期四):
36、第二场比赛3. 3月12日(星期六):第三场比赛4. 3月13日(星期日):第四场比赛5. 3月15日(星期二):第五场比赛 超级计算学院 人工智能 Chapter01 AboutAI AlphaGo战绩惊人战绩惊人实际上,目前AlphaGo已经成为最优秀的人工智能围棋程序。在与其他程序的对弈中,AlphaGo用一台机器就取得了500场的胜利,甚至有过 让对手4手后获胜的纪录。去年10月5日-10月9日,谷歌安排AlphaGo与欧洲围棋冠军Fan Hui(樊麾:法国国家围棋队总教练)闭门比赛,谷歌以5-0取胜。此次比赛的李世石是近10年来获得世界第一头衔最多的棋手,谷歌为此提供了100万美元作
37、为奖金。李世石谈到此次比赛时表示:”这是电脑首次在公平比赛中挑战人类专家选手,我很荣幸能参与进来。无论结果如何,这都是围棋史上的重要时刻。我听说Google DeepMind的人工智能出乎意料的强大,并且一直在优化,但至少这次我还是很自信能够取得胜利。” 超级计算学院 人工智能 Chapter01 AboutAI 3。 人工智能的研究领域。人工智能的研究领域。 The fields of Artificial Intelligence. 3.1 定理证明和计算机自动推理定理证明和计算机自动推理Alen Newell and Herbert Simon(Carnegie Tech)Logic T
38、heorist 吴文俊方法吴文俊方法 超级计算学院 人工智能 Chapter01 AboutAI 3。2 机器学习机器学习学习是少数高级生物具有的高级智能行为学习是少数高级生物具有的高级智能行为机器学习是人造系统有了自我完善能力机器学习是人造系统有了自我完善能力以前的机械系统,无论它有多么复杂巧妙,以前的机械系统,无论它有多么复杂巧妙, 他都他都只能以程序化的方式进行固定的操作,其能力和水只能以程序化的方式进行固定的操作,其能力和水平不会有任何的改进。而具有学习能力的计算机系平不会有任何的改进。而具有学习能力的计算机系统,其以后的水平如何,连设计者都难以预料。统,其以后的水平如何,连设计者都难
39、以预料。机器学习加快了学习的进程机器学习加快了学习的进程机器学习的延续性机器学习的延续性机器学习的快速传播性机器学习的快速传播性 超级计算学院 人工智能 Chapter01 AboutAI 3 3。3 3 自然语言理解自然语言理解 似乎很简单但实质上是非常困难的问题似乎很简单但实质上是非常困难的问题 日常的语言交流包含着很复杂的编码和解码过程日常的语言交流包含着很复杂的编码和解码过程 与人们的知识与推理能力有密切的关系与人们的知识与推理能力有密切的关系, , 与某些人与某些人 谈话很费力,谈话很费力, 但与思维敏捷的人谈话只要几句话但与思维敏捷的人谈话只要几句话 甚至几个字就可以了甚至几个字就
40、可以了 超级计算学院 人工智能 Chapter01 AboutAI 超级计算学院 人工智能 Chapter01 AboutAI 3.4专家系统3.5 机器人3.6 神经网3.7 遗传算法3.8 智能规划3.9 诊断3.10 智能代理 超级计算学院 人工智能 Chapter01 AboutAI 专家系统专家系统 专家系统是人工智能中应用最广泛的领域 专家系统是依赖于知识的人工智能系统 知识就是力量 最早的专家系统 DENDRAL(Lindsay, 1980, standford university), 根据化学分子式和光谱结构信息推断有机分子的结构, 应用在全世界的化学和药学实验中。MYCIN
41、(Buchanan and Shortliff, 1984)奠定了当代专家系统方法的基础, 诊断脊髓脑膜炎和血液传染病, 并开医疗处方。PROSPECT. 超级计算学院 人工智能 Chapter01 AboutAI 专家系统存在的问题:专家系统存在的问题: 1. 缺乏问题的结构知识。缺乏问题的结构知识。 2. 缺乏鲁棒性和灵活性缺乏鲁棒性和灵活性 3. 不能对诊断提供解释不能对诊断提供解释 4. 专家知识需要积累过程专家知识需要积累过程 5.自我获取知识能力差自我获取知识能力差 超级计算学院 人工智能 Chapter01 AboutAI 4. 人工智能的应用人工智能的应用 The Applic
42、ations of Artificial Intelligence 4.1 知识发现和数据挖掘知识发现和数据挖掘 The knowledge discovery and data mining BACON system 4.2 嵌入式智能系统嵌入式智能系统 Embeded intelligence system 只要是在一般的应用系统中嵌入一点小的智能系统,只要是在一般的应用系统中嵌入一点小的智能系统, 就会使系统的功能得到很大的提高。就会使系统的功能得到很大的提高。 例如,例如, 智能汉字输入智能汉字输入 语音输入语音输入 自动翻译自动翻译 手写识别和输入手写识别和输入 人脸识别,人脸识别,
43、 气味识别。气味识别。 超级计算学院 人工智能 Chapter01 AboutAI 43 智能规划的应用智能规划的应用 人工智能是计算机研究中应用性很强的一个领域. 例如, 在工厂作业调度规划问题中(Job shop scheduling), 就是要考虑在有限的加工资源(车床,刨床,钻床)的情况下, 根据已知的工件的加工顺序要求对整个车间的生产作出安排, 使得加工完所有工件所需的时间尽可能的少, 每台机床的等待时间尽可能的短。 另一个典型的工厂作业调度规划问题是考虑在有限辆的货运汽车的前提下, 在不同的地点之间运送货物. 规划的输出是一张车辆运转计划表, 使得汽车尽可能地满载运输, 空车运行情
44、况尽可能地少, 车辆闲置的情况尽可能地少, 这当然也会给运输公司带来可观的效益. 超级计算学院 人工智能 Chapter01 AboutAI 除了在现代化的工厂里有大量人工智能应用外, 人工智能的另一个重要应用领域是宇宙航行. 这方面应用的一个具体例子是哈勃空间望远镜(HST, Hubble Space Telescope)的修复, 在修复过程中,地面人员不断得到关于HST能作什么、不能作什么的最新信息, 然后对修复工作作出规划, 从而使HST恢复了正常观测能力4. 正因为智能规划有如此广泛的应用,所以人工智能专家McDermott指出“智能规划调度问题大量地出现在工业领域,规划质量的改进,哪
45、怕是一点小小的改进,都会节约大量的时间,带来上百万美元的效益”5. 超级计算学院 人工智能 Chapter01 AboutAI Intelligent planning and scheduling problems exist in the field of industry, improve the quality of the plan, even if it is a little improvement, will save a lot of time to bring millions of dollars in benefits. McDermott 1995 超级计算学院 人工
46、智能 Chapter01 AboutAI 6.1 智能规划在航空航天上的应用智能规划在航空航天上的应用 为说明智能规划在航空航天上的应用, 我引用一篇“宇航学报“上的文章, 文章的题目是”航天器自主运行技术的进展“, 这是宇航专业的文章, 可是它大部分讲的是计算机技术, 特别是智能规划技术和基于模型诊断技术。文献9的摘要如下摘要:阐述了航天器自主运行的概念、目标和任务。对自主运行和传统测控方式进行了比较。最后重点介绍了航天器自主运行技术的进展情况。文章分4个部分介绍自主运行技术。首先介绍了2种自主运行体系结构,它们是自主运行各种功能集成的基础。第第2部分介绍了部分介绍了2种智能规划与调度技术。
47、第种智能规划与调度技术。第3部分介绍了基于模型的故障部分介绍了基于模型的故障诊断与系统重构技术。诊断与系统重构技术。第4部分介绍了有效载荷数据自主处理的进展情况。最后进行了总结并介绍了与自主运行相关的其他技术。 关键词:航天器;自主智能控制;智能规划与调度;故障诊断;数据处理中图分类号:V423文献标识码:A文章编号:1000-1328(2003)01-0017-06 超级计算学院 人工智能 Chapter01 AboutAI AI, 导弹导弹控制与高性能计算机控制与高性能计算机A launch control center (LCC), in the United States, is th
48、e main control facility for intercontinental ballistic missiles(ICBMs). A launch control center monitors and controls missile launch facilities. From a launch control center, the missile combat crew can monitor the complex, launch the missile, or relax in the living quarters (depending on the ICBM s
49、ystem). The LCC is designed to provide maximum protection for the missile combat crew and equipment vital to missile launch. Missile silos are common across the midwestern United States, and over 450 missiles remain in US Air Force service. United States 超级计算学院 人工智能 Chapter01 AboutAI Due to modern c
50、onventional weapons, missile launch control centers are becoming more rare in the US, and it is expected that the number of missiles will lower to around 450 Minuteman 3s. However, the USAF remains alert to any outbound threats and is prepared to launch their missiles in the event of nuclear war. 超级