ARTIFICIAL INTELLIGENCE 2018/2019 Semester 1 Introduction: Chapter 1
CS410 Course home page: http://bcmi.sjtu.edu.cn/ai/ schedule, lecture notes, tutorials, assignment, grading, office hours, etc. Textbook: S. Russell and P. Norvig Artificial Intelligence: A Modern Approach Prentice Hall, 2010, Second Edition Lecturer: Liqing Zhang Grading: Assignment +Class Test(40%), Projects (20%), Final report (40%)
Course overview Introduction and Agents (chapters 1,2) Search (chapters 3,4,5,6) Logic (chapters 7,8,9) Planning (chapters 11,12) Uncertainty (chapters 13,14) Learning (chapters 18,20) Natural Language Processing (chapter 22,23)
Outline Course overview What is AI? A brief history The state of the art
What is AI? Artificial Intelligence (AI) Intelligent behavior Intelligent behavior in artifacts “Design computer programs to make computers smarter” “Study of how to make computers do things at which, at the moment, people are better” Intelligent behavior Perception, reasoning, learning, decision, communicating, acting in complex environments Long term goals of AI Develop machines that do things as well as humans can or possibly even better Understand intelligent behaviors
What is AI? Can machines think? “Can” “Machine” Depend on the definitions of “machine”, “think”, “can” “Can” Can machines think now or someday? Might machines be able to think theoretically or actually? “Machine” E6 Bacteriophage: Machine made of proteins Searle’s belief Thinking can occur only in very special machines – living ones made of proteins
What is AI? Views of AI fall into four categories: Thinking humanly Thinking rationally Acting humanly Acting rationally The textbook advocates "acting rationally"
Acting humanly: Turing Test Turing (1950) "Computing machinery and intelligence": "Can machines think?" "Can machines behave intelligently?" Operational test for intelligent behavior: the Imitation Game Predicted that by 2000, a machine might have a 30% chance of fooling a lay person for 5 minutes Suggested major components of AI: knowledge, reasoning, language understanding, learning
Thinking humanly: cognitive modeling 1960s "cognitive revolution": information-processing psychology Requires scientific theories of internal activities of the brain How to validate? Requires 1) Predicting and testing behavior of human subjects (top-down) 2) Direct identification from neurological data (bottom-up) Both approaches (Roughly, Cognitive Science and Cognitive Neuroscience), are now distinct from AI
Thinking rationally: "laws of thought" Aristotle: what are correct arguments/thought processes? Several Greek schools developed various forms of logic: notation and rules of derivation for thoughts; may or may not have proceeded to the idea of mechanization Direct line through mathematics and philosophy to modern AI Problems: Not all intelligent behavior is mediated by logical deliberation What is the purpose of thinking? What thoughts should I have?
Acting rationally: rational agent Rational behavior: doing the right thing The right thing: that which is expected to maximize goal achievement, given the available information Doesn't necessarily involve thinking – e.g., blinking reflex – but thinking should be in the service of rational action
Rational agents An agent is an entity that perceives and acts Abstractly, an agent is a function from percept histories to actions: [f: P* A] For any given class of environments and tasks, we seek the agent (or class of agents) with the best performance Remark: computational limitations make perfect rationality unachievable design best program for given machine resources
AI prehistory Philosophy Logic, methods of reasoning, mind as physical system foundations of learning, language, rationality Mathematics Formal representation and proof algorithms, computation, (un)decidability, (in)tractability, probability Economics Utility, Decision theory Neuroscience Physical substrate for mental activity Psychology Phenomena of perception and motor control, experimental techniques Computer Building fast computers engineering Control theory Design systems that maximize an objective function over time Linguistics Knowledge representation, grammar
Abridged history of AI 1943 McCulloch & Pitts: Boolean circuit model of brain 1950 Turing's "Computing Machinery and Intelligence" 1956 Dartmouth meeting: "Artificial Intelligence" adopted 1952—69 Look, Ma, no hands! 1950s Early AI programs, including Samuel's checkers program, Newell & Simon's Logic Theorist, Gelernter's Geometry Engine 1965 Robinson's complete algorithm for logical reasoning 1966—73 AI discovers computational complexity Neural network research almost disappears 1969—79 Early development of knowledge-based systems 1980-- AI becomes an industry 1986-- Neural networks return and became popular 1987-- AI becomes a science 1995-- The emergence of intelligent agents
State of the Art Deep Blue defeated the reigning world chess champion Garry Kasparov in 1997 Proved a mathematical conjecture (Robbins conjecture) unsolved for decades (1996 by W. McCune) No hands across America (driving autonomously 98% of the time from Pittsburgh to San Diego) During the 1991 Gulf War, US forces deployed an AI logistics planning and scheduling program that involved up to 50,000 vehicles, cargo, and people NASA's on-board autonomous planning program controlled the scheduling of operations for a spacecraft Proverb solves crossword puzzles better than most humans
Computer Sci. and Brain Sci. Information Processing in Digital Computer Computing based on Logic CPU and Storage: Separated Data Processing & Storage: Simple Intelligent Information Processing: Complicated and Slow Cognitive capability: Weak Information Process Mode: Logic – Information – Statistics Information Processing in the Brain Computing based on Statistics CPU and Storage: Unified Data Processing & Storage: Unknown Intelligent Information Processing: Simple and Fast Cognitive capability: Strong Information Process Mode: Statistics －concepts－logic
Visual Information Processing Fig.2.29 ‘Where’: the motion and spatial location ‘What’: the detailed features, form, and object identity
Biological Neurons
Challenges
Human Vision (1)
Human Vision (3)
Relax for a while Test：How many human faces in the picture？
人工智能简史 2016 Alpha GO 2012 深度卷积网络, Le Kun 2011 IBM Watson 2005 深度学习, Hinton 1996 Robbins 猜想 1986-神经网络 1965 Robinson‘s 完备算法 1957 感知器, Rosenblatt 1956人工智能定义Dartmouth workshop 1943神经元模型 McCulloch&Pitts
类脑模型进展 CNN, 1998, 2012 RNN, 1982, 2013 Deep learning, 2005, 2009 S. Amari 1967 S. Grossberg Deep learning, 2005, 2009 F. Rosenblatt Hopfield Neural network（1982） The multilayer perceptron(1972) Fukushima, Neocognitron(1980) The perceptron algorithm （1957） McCulloch－Pitts′ neuron model （ 1943 ）
深度典型实例 – 图像标注
IBM再次赢得“人机大战” IBM 高调推出超级计算机 Watson，目标是建造一个能与人类回答问题能力匹敌的计算系统，在比赛中，参赛者必须要回答一系列的问题，主要涉及历史，文学，政治，电影，流行文化和科学。 这要求计算机具有足够的速度、精确度和置信度，并且能使用人类的自然语言回答问题。 挑战——回答 Jeopardy 比赛中的题目需要分析人类语言中微妙的含义、讽刺口吻、谜语等，这些通常是人类擅长的方面，一直以来计算机在这方面毫无优势可言。 在美国最受欢迎的智力竞猜节目播放的2011年2月14日-2月16日期间，IBM超级电脑Watson其中击败了两名人类选手，最终获得胜利。
实例分析：AlphaGo 利用深度学习、强化学习，拓展 了机器 “直觉感知”“棋局推理（ 全局获胜机会如何）”和“新颖落 子（想人所不敢想）”等能力， 并将记忆人类棋局和自我博弈积 累棋局结合了起来。 全新的复杂问题求解技术路线 学到专家群体的智慧，而不仅仅 是个体专家的智慧 引入深度网络学习棋盘布局 模式 学习专家群体的价值网络 利用强化学习价值网络 利用随机搜索和博弈算法
金融智能：信用卡欺诈检测算法 传统人工之智能方法 新一代人工智能方法 … … 监督学习与无监督学习 数据驱动 统计分析 专家系统方法 （决策树） 神经网络方法 异常检测方法 模糊逻辑方法 … … 新一代人工智能方法 数据驱动 利用深度学习模型正常交易模式和 欺诈模式 利用强化学习方式学习欺诈损失和 检测成本 利用学习正常交易模式的演变，预 测新型可疑欺诈交易
人工智能基本问题 让机器能像人那样理解、思考和学习。AI进步的动力不仅来 自于内部驱动，更来自于信息环境与社会需求等外部驱动。 问题求解与搜索 知识表示与逻辑推理 不确定推理、机器学习 自然语言、计算机视觉、语音识别 人机交互、场景理解、对话交互 多智能体/群体智能 博弈与对策
人工智能：机遇与挑战 场景理解??? 信息感知 知识推理 AI关键技术取得突破 典型应用 深度学习理论重大进展 关键技术突飞猛进 语音识别、图像模式识别 自然语言处理、知识工程 …… 典型应用 计算机围棋(Alpha Go, Learning from Experience/Rules, Value, Policy) 自动驾驶 (Autonomous Vehicles) 医学影像自动诊断（Medical Imaging Diagnosis） 智能客服/智能音箱 …… 重大理论问题与核心技术有待突破（自主学习、知识表示、不确定性推 理、场景理解、智能交互、执行与反馈、群体博弈) 场景理解??? 信息感知 知识推理
Intelligent Systems
Home work To write a short report on your personal interests in the field of AI.
围棋课程设计报告提要 项目任务分工简介 课题实现的技术路线 采用方法简介（除了本课程的算法，是否还采用了其他算法） 实验结果分析 结论与感想 每一个同学在项目分工、工作量分配 课题实现的技术路线 采用方法简介（除了本课程的算法，是否还采用了其他算法） 实验结果分析 结论与感想