Introduction to Artificial Intelligence Unit 10 Communication Course

Introduction to Artificial Intelligence – Unit 10 Communication Course 67842 The Hebrew University of Jerusalem School of Engineering and Computer Science Academic Year: 2008/2009 Instructor: Jeff Rosenschein (Chapter 22, “Artificial Intelligence: A Modern Approach”)

Real Language § Real human languages provide many problems for NLP: § ambiguity § anaphora § indexicality § vagueness § discourse structure § metonymy § metaphor § noncompositionality 2

Ambiguity § § § Beach closing down last year Squad helps dog bite victim Helicopter powered by human flies American pushes bottle up Germans I ate spaghetti with meatballs § § with salad with abandon with a fork with a friend § Ambiguity can be lexical (polysemy), syntactic, semantic, referential 3

Ambiguity resolved in speech § The sentence: “I never said she stole my money. ” can have seven different meanings depending on which word is stressed § I never said she stole my money. 4

Anaphora § Using pronouns to refer back to entities already introduced in the text: § After Mary proposed to John, they found a preacher and got married. § For the honeymoon, they went to Hawaii. § Mary saw a ring through the window and asked John for it. § Mary threw a rock at the window and broke it. 5

Indexicality § Indexical sentences refer to utterance situation (place, time, S/H, etc. ): § I am over here. § Why did you do that? 6

Metonymy § Using one noun phrase to stand for another: § I’ve read Shakespeare. § Chrysler announced record profits. § The ham sandwich on Table 4 wants another beer. 7

Metaphor § “Non-literal” usage of words and phrases, often systematic: § I’ve tried killing the process but it won’t die. Its parent keeps it alive. 8

Noncompositionality § § § basketball shoes baby shoes alligator shoes designer shoes brake shoes § § red book red pen red hair red herring § § § small moon large molecule mere child alleged murderer real leather artificial grass 9

What this Course Covered 10

Overall Structure § Intelligent Agents, What is AI? § Search § Knowledge Representation § Planning § Probabilistic Reasoning § Game Theory § Natural Language/Learning 11

More Detailed Structure § Introduction: what is AI? the Turing Test; History of AI; state of the art § Intelligent Agents: rationality, environments, agent structure § Search: breadth-first, depth-first, iterative deepening, bidirectional search, informed heuristic search, A*, heuristic functions, hill climbing, simulated annealing, Constraint satisfaction problems, backtracking search for CSPs, Adversarial search, games, minimax, alpha-beta pruning 12

More Detailed Structure 2 § Knowledge Representation: propositional logic; propositional inference, first-order logic; quantifiers; encoding of knowledge, inference in first-order logic, unification, forward chaining, backward chaining, resolution § Planning: planning with state-space search, partial order planning, planning graphs, planning with propositional logic, hierarchical task network planning, conditional planning, continuous planning, multiagent planning 13

AI: A Dynamic Field § There are many ways of categorizing approaches to problems in AI § Neat vs. Scruffy § Theoreticians vs. Experimentalists § Rule-based vs. data-based § Users of particular “tools” or “approaches” § POMDPs § Learning § And more… 14

What are the State-of-the-Art Research Topics? § IJCAI’ 09 meets in Pasadena, July 2009 § Session topics § Cognitive and Philosophical Foundations § Performance and Behavior Modeling in Games § Depth and Breadth First Search § Time Series/Activity Recognition § Diagnosis and Testing § Automated Reasoning § Unsupervised Learning I § Social Choice I: Manipulation § Search in Games 15

§ Plan Recognition § Ontology Matching and Learning § Spatial Reasoning § Semi-Supervised Learning I § Multimodal Interaction § Online Games § Distributed Constraint Satisfaction § Model-Based Diagnosis and Applications § Causality and Graphical Models § Transfer Learning § Word Sense Disambiguation § Recommender Systems § Satisfiability I: Extensions and Applications § Multiagent Planning and Learning § Robotics: Multirobot Planning 16

§ Preferences: Learning I § Search and Learning § Multiagent Resource Allocation § Argumentation I § Epistemic Logic § Semi-Supervised Learning II: Applications § HTN Planning § Coalitional Games § Unsupervised Learning II § Heuristic Search § Constraints I: Global Constraints § Logic Programming I § Mechanism Design § Reasoning about Action I § Clustering 17

§ Text Summarization & Understanding § Preferences: Learning II § Local and Anytime Search § Game Theory: Solution Concepts § Social Choice II: Voting § Constraints II § Optimal Planning § Description Logics I: Reasoning § Metric Learning § POMDPs II § Morphology and Counting § Vision & Robotics I: Novelty § Preferences: Graphical Models § Planning: Search Techniques § Vision & Robotics II 18

§ Social Choice III § Advances in A* Search § Contingent and Nondeterministic Planning § Activity and Goal Recognition § Reasoning about Action II § Parsing and Translation § Coalitions and Coordination § Learning: Dimensionality Reduction § Inference in Graphical Models § Games and Monte Carlo Search § Web Mining and Web Services § Negotiation and Commitment § Spatio-Temporal Reasoning/Distributed & Game. Theoretic KR § Learning Relational and Graphical Models 19

§ Kernel Methods § Natural Language Semantics § Musical Expression/Vision & Robotics III § Constraints III § Logic Programming II § Description Logics II: Query Answering § Auctions § Structure Learning § Markov Decision Processes § Satisfiability II § Description Logics III: Non-standard Reasoning § Argumentation II § Social Networks § Learning: Matrix Factorization § Reinforcement Learning 20

What are the AI Apps to Come? § Long-held dreams are coming true: § Language Translation § Speech Recognition § Mundane tasks made possible by learning from data: § Fare. Cast § What would we want a machine to do, that it can’t do now? § Autonomous Driving? § Teaching? 21

AI (and Software) Ethical Issues § When computers are programmed to take the § § § place of humans, where does liability reside? Is fast behavior unethical, when slow versions of the same behavior are ethical (e. g. , machine scanning of vast amounts of mortgage information, publically available, that would be much harder to analyze if done by a human)? Human-machine symbiosis – what crosses the line? Machine-machine behavior – is any behavior that is unethical for humans allowed for computers? Vice versa? 22