C463_01_introC463_01_introC463_01_introC463_01_intro
- 1. C463 / B551 C463 / B551 Artificial Intelligence Artificial Intelligence Dana Vrajitoru Dana Vrajitoru Introduction Introduction
- 2. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Course Outline Course Outline Introduction, definition, philosophy Introduction, definition, philosophy Intelligent agents Intelligent agents Logic, knowledge representation, reasoning Logic, knowledge representation, reasoning Fuzzy logic, probabilistic reasoning Fuzzy logic, probabilistic reasoning Planning, game playing, decision-making Planning, game playing, decision-making Expert systems Expert systems Machine learning Machine learning Genetic algorithms, neural networks, SOM Genetic algorithms, neural networks, SOM Elements of natural language processing. Elements of natural language processing.
- 3. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Artificial Intelligence Artificial Intelligence Definition Definition. The science of developing methods to . The science of developing methods to solve problems usually associated with human solve problems usually associated with human intelligence. intelligence. Alternate definitions: Alternate definitions: building intelligent entities or agents; building intelligent entities or agents; making computers think or behave like humans making computers think or behave like humans studying the human thinking through computational studying the human thinking through computational models; models; generating intelligent behavior, reasoning, learning. generating intelligent behavior, reasoning, learning.
- 4. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Questions Questions What do we call intelligence? What do we call intelligence? Examples of intelligent tasks. Examples of intelligent tasks. Can an artificial being ever be considered Can an artificial being ever be considered "alive"? What does it mean to be "alive"? "alive"? What does it mean to be "alive"?
- 5. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Natural Intelligence Natural Intelligence Definition. Definition. Intelligence Intelligence – inter ligare (Latin) – the capacity – inter ligare (Latin) – the capacity of creating connections between notions. of creating connections between notions. Wikipedia: the ability to solve problems. Wikipedia: the ability to solve problems. WordNet: the ability to comprehend; to understand and WordNet: the ability to comprehend; to understand and profit from experience. profit from experience. Complex use of creativity, talent, imagination. Complex use of creativity, talent, imagination. Biology - Intelligence is the ability to adapt to new Biology - Intelligence is the ability to adapt to new conditions and to successfully cope with life situations. conditions and to successfully cope with life situations. Psychology - a general term encompassing various mental Psychology - a general term encompassing various mental abilities, including the ability to remember and use what abilities, including the ability to remember and use what one has learned, in order to solve problems, adapt to new one has learned, in order to solve problems, adapt to new situations, and understand and manipulate one’s reality. situations, and understand and manipulate one’s reality. Nonlinear, non-predictable behavior. Nonlinear, non-predictable behavior.
- 6. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Visions of AI Visions of AI Systems that think like humans. Systems that think like humans. Systems that act like humans. Systems that act like humans. Systems that think rationally. Systems that think rationally. Systems that act rationally. Systems that act rationally. A distinction between being intelligent and A distinction between being intelligent and acting intelligently, and being like a human, acting intelligently, and being like a human, or solving similar problems (not necessarily or solving similar problems (not necessarily the same way). the same way).
- 7. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Thinking Humanly Thinking Humanly Cognitive science: modeling the processes of Cognitive science: modeling the processes of human thought. human thought. Through a set of experiments and computational Through a set of experiments and computational models, trying to build good explanations of models, trying to build good explanations of what we do when we solve a particular task. what we do when we solve a particular task. Relevance to AI: to solve a problem that humans Relevance to AI: to solve a problem that humans (or other living being) are capable of, it's good to (or other living being) are capable of, it's good to know how we go about solving it. know how we go about solving it. Early approaches tried to solve any problem Early approaches tried to solve any problem exactly the way a human would do. Now we exactly the way a human would do. Now we know that it's not the best approach. know that it's not the best approach.
- 8. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Acting Humanly Acting Humanly How do you distinguish intelligent behavior from How do you distinguish intelligent behavior from intelligence? intelligence? Turing test Turing test, by A. Turing, 1950: determining if a , by A. Turing, 1950: determining if a program qualifies as artificially intelligent by program qualifies as artificially intelligent by subjecting it to an interrogation along with a subjecting it to an interrogation along with a human counterpart. human counterpart. The program passes the test if a human judge The program passes the test if a human judge cannot distinguish between the answers of the cannot distinguish between the answers of the program and the answers of the human subject. program and the answers of the human subject. It hasn't been passed yet. It hasn't been passed yet. http://www.loebner.net/Prizef/loebner-prize.html http://www.loebner.net/Prizef/loebner-prize.html
- 9. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Thinking Rationally Thinking Rationally Systems capable of reasoning, capable of Systems capable of reasoning, capable of making logical deductions from a knowledge making logical deductions from a knowledge base. base. This requires some capacity to make logical This requires some capacity to make logical inferences, like " inferences, like "All humans are mortal; Socrates All humans are mortal; Socrates is a human; thus Socrates is mortal is a human; thus Socrates is mortal". ". Good news Good news: formal logic is easy to express as a : formal logic is easy to express as a program and its rules are clear. program and its rules are clear. Bad news Bad news: G : Gö ödel's incompleteness theorem and del's incompleteness theorem and SAT is NP-Complete. SAT is NP-Complete.
- 10. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Gödel's Theorem Gödel's Theorem At some point it was believed that one could At some point it was believed that one could prove anything using only logic, building a formal prove anything using only logic, building a formal system to describe the knowledge - system to describe the knowledge - Hilbert Hilbert. . K. K. Gödel Gödel proved in his proved in his Incompleteness Theorem Incompleteness Theorem that within any formal system, some statements that within any formal system, some statements that are true could not be proven using only that are true could not be proven using only formal logic based on the axioms of that system. formal logic based on the axioms of that system. What this means What this means: logic is a powerful and : logic is a powerful and necessary tool in automatic reasoning, but to necessary tool in automatic reasoning, but to make useful deductions one requires domain- make useful deductions one requires domain- specific knowledge. specific knowledge.
- 11. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru SAT – NP-Complete SAT – NP-Complete SAT SAT – satisfiability problem. Given a logical – satisfiability problem. Given a logical formula involving a set of Boolean variables, is formula involving a set of Boolean variables, is there a set of values for these variables such there a set of values for these variables such that the formula is true? that the formula is true? Relevance to AI: the problem of deciding if Relevance to AI: the problem of deciding if something is true in a given system (making a something is true in a given system (making a deduction) comes down to solving a particular deduction) comes down to solving a particular SAT problem. SAT problem. NP-complete NP-complete: there is no known polynomial : there is no known polynomial algorithm to solve this problem, but if we find algorithm to solve this problem, but if we find one for it, then we can solve any other NP one for it, then we can solve any other NP problem. For now a guaranteed solution is problem. For now a guaranteed solution is exponential exponential. .
- 12. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Acting Rationally Acting Rationally Many AI applications adopt the intelligent Many AI applications adopt the intelligent agent agent approach. approach. An An agent agent is an entity capable of generating is an entity capable of generating action. action. In AI a rational agent must be autonomous, In AI a rational agent must be autonomous, capable of perceiving its environment, capable of perceiving its environment, adaptable, with a given goal. adaptable, with a given goal. Most often the agents are small pieces of code Most often the agents are small pieces of code with a specific proficiency. The problem is solved with a specific proficiency. The problem is solved by combining the skills of several agents. by combining the skills of several agents.
- 13. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru History of AI History of AI 1943 1943 – W. McCulloch and W. Pitts designed the – W. McCulloch and W. Pitts designed the first neural network. M. Minsky and D. Edmonds first neural network. M. Minsky and D. Edmonds built the first one in 1951 at Princeton. built the first one in 1951 at Princeton. 1950 1950 – A. Turing, "Computing Machinery and – A. Turing, "Computing Machinery and Intelligence". Intelligence". 1956 1956 – J. McCarthy organized a workshop at – J. McCarthy organized a workshop at Darmouth where the name of AI was officially Darmouth where the name of AI was officially adopted for the field. adopted for the field. Early successes: the General Problem Solver Early successes: the General Problem Solver (puzzles), Geometry Theorem Prover, Samuel's (puzzles), Geometry Theorem Prover, Samuel's checkers player. checkers player. 1958 1958 – McCarthy invented Lisp. – McCarthy invented Lisp.
- 14. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru History of AI History of AI The early systems were successful on small The early systems were successful on small problems but failed on larger ones. problems but failed on larger ones. 1958 1958 – Friedberg's machine evolution (now – Friedberg's machine evolution (now better known as hill-climbing) using mutations; it better known as hill-climbing) using mutations; it failed to find good solutions. failed to find good solutions. 1966 1966 – a commission reports on the failing of – a commission reports on the failing of machine translation and all funding to such machine translation and all funding to such projects is ceased. projects is ceased. 1969 1969 – Minsky and Papert, Perceptrons, proved – Minsky and Papert, Perceptrons, proved that they could learn anything they could that they could learn anything they could represent, but there was not much they could represent, but there was not much they could represent. represent.
- 15. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru History of AI History of AI Knowledge-based systems – that contain Knowledge-based systems – that contain domain-specific knowledge giving them more domain-specific knowledge giving them more problem-solving power – problem-solving power – Expert Systems Expert Systems. The . The industry adopted them on a relatively large industry adopted them on a relatively large scale, but many such projects failed. scale, but many such projects failed. More recent developments combine AI methods More recent developments combine AI methods with strategies from other fields. with strategies from other fields. Although the initial ambition of AI seems a Although the initial ambition of AI seems a distant goal at most, many methods have been distant goal at most, many methods have been developed that are used in most areas of CS. developed that are used in most areas of CS.
- 16. Successes in AI Successes in AI 1975 – Meta-Dendral learning program finds new rules in 1975 – Meta-Dendral learning program finds new rules in spectral chemistry. spectral chemistry. 1978 – Herb Simon wins the Nobel Prize in Economics 1978 – Herb Simon wins the Nobel Prize in Economics for his theory of bounded rationality. for his theory of bounded rationality. 1979 - The Stanford Cart, built by Hans Moravec, the 1979 - The Stanford Cart, built by Hans Moravec, the first computer-controlled autonomous vehicle. first computer-controlled autonomous vehicle. 80s – neural networks with backpropagation algorithm 80s – neural networks with backpropagation algorithm become popular, evolutionary computation become popular, evolutionary computation 1997 – Deep Blue beats G. Kasparov, first Robo-Cup. 1997 – Deep Blue beats G. Kasparov, first Robo-Cup. 2000 – Interactive robots commercially available, Kismet 2000 – Interactive robots commercially available, Kismet (MIT), robots used for real applications. (MIT), robots used for real applications. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru
- 17. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Related Fields Related Fields Philosophy – knowledge, mind, logic Philosophy – knowledge, mind, logic Mathematics - formal rules, logic, probability, Mathematics - formal rules, logic, probability, algorithms algorithms Economics – decision making, maximizing the Economics – decision making, maximizing the outcome, game theory outcome, game theory Neuroscience – understanding how the brain Neuroscience – understanding how the brain works works Psychology – How do animals and humans think Psychology – How do animals and humans think and act? and act? Cybernetics – control theory Cybernetics – control theory Linguistics – understanding the natural language Linguistics – understanding the natural language
- 18. Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru Main Areas of AI Main Areas of AI Autonomous planning and scheduling Autonomous planning and scheduling Decision making Decision making Machine learning, adaptive methods Machine learning, adaptive methods Biologically inspired algorithms Biologically inspired algorithms Game playing Game playing Autonomous control, robotics Autonomous control, robotics Natural language processing Natural language processing
- 19. Relevant Publications Relevant Publications Machine Learning – journal, Springer. – journal, Springer. ACM SIGART special interest group, SIGEVO. ACM SIGART special interest group, SIGEVO. AAAI society, annual conference, journal. AAAI society, annual conference, journal. International Joint Conference on Artificial International Joint Conference on Artificial Intelligence (IJ-CAI), bi-annual. Intelligence (IJ-CAI), bi-annual. GECCO – SIGEVO conference on evolutionary GECCO – SIGEVO conference on evolutionary computation. computation. IEEE Transactions on Pattern Analysis and Mac hine Intelligence Artificial Intelligence – D. Vrajitoru Artificial Intelligence – D. Vrajitoru