2 Introduction to AI: What is Intelligence? Intelligence, taken as a whole, consists ofthe following skills:-1. the ability to reason2. the ability to acquire and apply knowledge3. the ability to manipulate andcommunicate ideas
3 Introduction to AI: Definitions of AI " ... the science of making machines do things that would require intelligence if done by humans"- Marvin MinskyAI is the part of computer science concerned with designing intelligent computer systems -E. FeigenbaumSystems that can demonstrate human-like reasoning capability to enhance the quality of life and improve business competitiveness Japan-S’pore AI Centre
4 Introduction to AI: An Intelligent Entity INPUTSINTERNALPROCESSESHas knowledgeSenses environmentHas understanding/intentionalitySeeHearTouchTasteSmellCan ReasonExhibits behaviourOUTPUTS
5 What is Artificial Intelligence Different definitions due to different criteriaTwo dimensions:Thought processes/reasoning vs. behavior/actionSuccess according to human standards vs. success according to an ideal concept of intelligence: rationality.Systems that think like humansSystems that think rationallySystems that act like humansSystems that act rationally
6 Systems that act like humans When does a system behave intelligently?Turing (1950) Computing Machinery and IntelligenceOperational test of intelligence: imitation gameTest still relevant now, yet might be the wrong question.Requires the collaboration of major components of AI: knowledge, reasoning, language understanding, learning, …
7 Introduction to AI: Turing’s Test - Part 1 Which one’s the man?ABPart 1 - Woman, Man & Judge.
8 Introduction to AI: Turing’s Test - Part 2 Part 2 - Woman, Machine & Judge.If the computer succeeds in fooling the judge then it has managed to exhibit a human level of intelligence in the task of pretending to be a woman, the definition of intelligence the machine has shown itself to be intelligent.Which one’s the computer?AB
9 Systems that act like humans Andrew Hodges.Alan Turing, the enigmaAvailable at amazon.co.ukProblem with Turing test: not reproducible, constructive oramenable to mathematical analysis.
10 Systems that think like humans How do humans think?Requires scientific theories of internal brain activities (cognitive model):Level of abstraction? (knowledge or circuitry?)Validation?Predicting and testing human behaviorIdentification from neurological dataCognitive Science vs. Cognitive neuroscience.Both approaches are now distinct from AIShare that the available theories do not explain anything resembling human intelligence.Three fields share a principal direction.
11 Systems that think like humans Some references;Daniel C. Dennet. Consciousness explained.M. Posner (edt.) Foundations of cognitive scienceFrancisco J. Varela et al. The Embodied MindJ.-P. Dupuy. The mechanization of the mind
12 Systems that think rationally Capturing the laws of thoughtAristotle: What are ‘correct’ argument and thought processes?Correctness depends on irrefutability of reasoning processes.This study initiated the field of logic.The logicist tradition in AI hopes to create intelligent systems using logic programming.Problems:Not all intelligence is mediated by logic behaviorWhat is the purpose of thinking? What thought should one have?
13 Systems that think rationally A reference;Ivan Bratko, Prolog programming for artificial intelligence.
14 Systems that act rationally Rational behavior: “doing the right thing”The “Right thing” is that what is expected to maximize goal achievement given the available information.Can include thinking, yet in service of rational action.Action without thinking: e.g. reflexes.Two advantages over previous approaches:More general than law of thoughts approachMore amenable to scientific development.Yet rationality is only applicable in ideal environments.Moreover rationality is not a very good model of reality.
15 Systems that act rationally Some references;Michael Wooldridge. Reasoning about rational agents.
16 Foundations of AIDifferent fields have contributed to AI in the form of ideas,viewpoints and techniques.Philosophy: Logic, reasoning, mind as a physical system, foundations of learning, language and rationality.Mathematics: Formal representation and proof algorithms, computation, (un)decidability, (in)tractability, probability.Psychology: adaptation, phenomena of perception and motor control.Economics: formal theory of rational decisions, game theory.Linguistics: knowledge representation, grammar.Neuroscience: physical substrate for mental activities.Control theory: homeostatic systems, stability, optimal agent design.
17 Introduction to AI: The Age of Intelligent Machines 1st Industrial Revolution: the Age of Automation: Machines extend & multiply man's physical capabilities2nd Industrial Revolution: the Age of Info Tech: Machines extend & multiply man's mental capabilitiesInformation & Knowledge Revolution: the Age of Knowledge Technology "..working smarter, not harder." How do we make our systems smarter? - by building in intelligence
18 Introduction to AI: Behaviorist's View on Intelligent Machines Many scientists believe that only things that can be directly observed are “scientific”Therefore if a machine behaves “as if it were intelligent” it is meaningless to argue that this is an illusion.Turing was of this opinion and proposed the “Turing Test”This view can be summarized as:“If it walks like a duck, quacks like a duck and looks like a duck - it is a duck”
19 Introduction to AI: History of AI Important research that laid the groundwork for AI:s: formal grammar & language theoriess: formalisation of reasoning (predicate calculus and propositional logic)s: Cybernetics - communication in man andmachine1950s: reality of digital computers (Mark I, ENIAC, EDVAC and UNIVAC)Others: Information Theory, Neurological Theories, Boolean Algebra, etc.
20 Introduction to AI: History of AI (cont’d) Basic philosophy is recorded since ancient GreeceEarly push after computer discovered (50's): Connectionist (neural net) vs. Symbolist/Logicist (AI)recognised as the official beginning of AI - The Dartmouth Summer WorkshopThe 1950s was also noted for chess playing programs, machine translation, automatic theorem provers, Chomsky generative grammars and LISPCMU, Stanford, and IBMEarly successes and enthusiasm - neural learning, theorem provers, problem solvers (GPS), game players, etc.
21 Introduction to AI: History of AI (cont’d) DartmouthConferenceDegreeofMotivationNewTechnologySupportJapan 5thGenerationComputerAIWinter19481970s - 80smid-1980sTimeAdapted from:Joe Carter (Andersen Consulting, 1988)Oliver Tian (Andersen Consulting, 1989)
22 Introduction to AI: Examples of AI systems RobotsChess-playing programVoice recognition systemSpeech recognition systemGrammar checkerPattern recognitionMedical diagnosisSystem malfunction rectifierGame PlayingMachine TranslationResource SchedulingExpert systems (diagnosis, advisory, planning, etc)Machine learningIntelligent interfaces
23 AI Case Study 1. Adaptive neural networks for intrusion detection My Ph.D. thesis topic, where I first introduced neural network techniques to aid allow information systems to learn new attacks and therefore adapt to changes in the threat environment
24 AI Case Study 2. Robot World Cup - Robocup The Robocup Competition pits robots (real and virtual) against each other in a simulated soccer tournament. The aim of the RoboCup competition is to foster an interdisciplinary approach to robotics and agent-based AI by presenting a domain that requires large-scale co-operation and coordination in a dynamic, noisy, complex environment. Common AI methods used are variants of neural networks and genetic algorithms.Pacific Rim International Conference on AI and Robocup Pacific Rim Series 1998, November 1998 NUS Multi-purpose Sports Hall.
26 Introduction to Philosophical Aspects of AI How is it possible for a physical thing--a person, an animal, a robot--to extract knowledge of the world from perception and then exploit that knowledge in the guidance of successful action? That is a question with which philosophers have grappled for generations.It could also be taken to be one of the defining questions of Artificial Intelligence.AI is, in large measure, philosophy. It is often directly concerned with instantly recognisable philosophical questions: What is mind? What is meaning? What is reasoning, and rationality? What are the necessary conditions for the recognition of objects in perception? How are decisions made and justified?
27 Introduction to Philosophical Aspects of AI (cont) Philosophy has a more direct relation to AI than it has to other sciences.Both subjects require the formalization of common sense knowledge.The subject of artificial intelligence may be seen as beginning with Turing's paper Computing Machinery and Intelligence (1950).
28 Philosophical Aspects of AI (cont) In fact, much AI already builds on work by philosophers. An obvious example is the use of speech act theory, developed originally by philosophers such as John Austin, John Searle and Paul Grice.There are also various uses of specialized logics, e.g. deontic logic, epistemic logic, and modal logics, originally developed by philosophers in an attempt to clarify concepts like `permission' and `obligation' (deontic logic), `knows' and `believes' (epistemic logic), and `necessarily' and `possibly' (modal logic).These contributions from philosophy are not passively accepted in AI: putting them to use in designing working systems often reveals shortcomings and suggests further development.
29 Introduction to Philosophical Aspects of AI (cont) Older contributions from philosophy includes Kant's proof in Critique of Pure Reason that learning from experience was impossible without some sort of prior (innate) conceptual apparatus.Another was Frege's heroic (but unsuccessful) attempt a century ago to show that all arithmetical concepts could be reduced to logical concepts and all arithmetical knowledge could be derived from logical axioms and rules. This led him to a number of extremely important results, including the first ever accurate analysis of the role of variables in mathematical expressions, discovery of the notion of higher order functions and invention of predicate calculus.
30 Introduction to Philosophical Aspects of AI (cont) This led (via work by Russell, Church and others) to lambda calculus, type theory, and other important notions in computer science and formalisms for AI.More recently the old philosophical controversy about varieties of forms of representations which has become a topic of active AI research.Another recent development is recognition of deep connections between the AI task of understanding what sort of knowledge an intelligent system requires and the older philosophical activities of metaphysics, especially what Strawson described as `descriptive metaphysics', including ontology, the attempt to characterize in a systematic way what exists.
31 Strong AI vs. Weak AI Terms introduced by John Searle. Strong AI: Thinking is just the manipulation of formal symbols; the mind is to the brain as the program is to the hardware; an properly programmed computer is the mind.Weak AI: Computer can teach us useful things about minds and brains, but they do not have minds. They can simulate mental activity and is merely a tool.