Artificial Intelligence Introduction to AI Dr. Muhammad Awais Outline • • • • • What is Artificial Intelligence (AI)? Foundation of AI History of AI Stat of Art Summary What is Artificial Intelligence (AI)? • • • • • Definitions Turing Test approach Cognitive modeling approach Thinking rationally Acting rationally Definitions • • • • Systems that think like humans Systems that think rationally Systems that act like humans Systems that act rationally Think human • • • “The exciting new effort to make computers think… machines with minds, in the full and literal sense” (Haugeland, 1985) “The automation of activities that we associate with human thinking, activities such as decision/making, problem solving, learning …” (Bellman, 1978) Cognitive modeling • • An approximation to animal cognitive processes (predominantly human) for the purposes of comprehension and prediction A cognitive process corresponds to a mental process that may include attention, memory, producing and understanding language, learning, reasoning, problem solving, and decision making. Act human • “The art of creating machines that perform functions that require intelligence when performed by people.” (Kurzweil, 1990) • The study of how to make computers do things at which, at the moment, people are better (Rich and Knight, 1991) Act human: Turing test • Predicted that by 2000, a machine might have a 30% chance of fooling a lay person for 5 minutes • Anticipated all major arguments against AI in following 50 years • Suggested major components of AI: knowledge, reasoning, language understanding, learning • Requirement of Turing test – – – – – – Natural language processing Knowledge representation Automated reasoning Machine learning Computer vision Robotics Think rational • • • “The study of mental faculties (abilities) through the use of computation models.” (Charniak and McDermott, 1985) “The study of computations that make it possible to perceive, reason, and act.” (Winston, 1992) Think-rational initiated the field of Logic (irrefutable reasoning) Example: “Socrates is a man; All men are mortal; Therefore, Socrates is mortal” (Syllogisms : inference of an output (proposition : a declarative sentence) from two or inputs (premises) Act rational • “Computational Intelligence is the study of the design of intelligent agents.” (Poole et al., 1990) • “AI … is concerned with intelligent behavior in artifacts”(robots). (Nilsson, 1998) Act rational • Agent : Not a mere computer program but • • • Autonomous operation Environment perception Adaption, etc 1. Best outcome in certain environment 2. Best expected outcome in uncertain environment Environment Perception Re/Action Environment Nao: http://en.wikipedia.org/wiki/Nao_(robot) Definitions Foundation of AI • • • • • • • • Philosophical contributions Mathematics Economics Neuroscience Psychology Computer engineering Control theory and Cybernetics Linguistics Philosophical contributions • Questions • • • • How does the mental mind arise from a physical brain ? Where does knowledge come from ? How does knowledge lead to action ? Contributions • Aristotle : Formulation of set of laws, e.g., syllogisms • • • A syllogism is a form of logical reasoning that joins two or more premises to arrive at a conclusion • Premises : All men are mortal • Premises : Aristotle is a man • Conclusion : Aristotle is mortal . Thomas Hobes : Reasoning by numerical computation . Leonardo da Vinci : Design of a mechanical calculator . Philosophical contributions • Dualism – part of the human mind (or soul or spirit) that is outside of nature, exempt from physical laws • Materialism – – Brain’s operation according to the laws of physics constitutes the mind. Free will is simply the way that the perception of available choices appears to the choosing entity • Empiricism – Nothing is in the understanding, which was not first in the senses.” • Induction – General rules are acquired by exposure to repeated associations between their elements Mathematical contributions • Questions • • • • • What are the formal rules to draw valid conclusions ? What can be computed ? How does we reason with uncertain information ? How does knowledge lead to action ? Contributions • • • • • • Euclid`s algorithm : Greatest common denominators computation David Hilbert : Proposition of problem for mathematicians Kurt Goedel : Incompleteness theorem (statements whose truth can not be established by an algorithm) Cobham, Edmonds : Notion of Intractability (time to solve grows exponentially with the size of problem instance) Steven Cook, Richard Karp : NP-completeness (Non deterministic polynomial time) Fermat, Pascal, Bernoulli, Laplace,… : Probability Economics contributions • Questions • • • • so as to How should we make decisions to maximize payoff ? How should we do this when other may not go along ? How should we do this when the payoff is far in the future ? Contributions • • • Leon Walras : Utility (preference of A over B) Pascal, Bernoulli : Decision theory (utility theory + probability theory ) Richard Bellman : Markov Decision Process (MDP) . . . Neuroscience contributions • Questions • • How do brains process information ? Contributions • • • Paul Brocas : speech production in left hemisphere (Broca’s area) Hans Berger : Measurement of intact brain activity Camillo Golgi : Observation of individual neurons . . . Psychology contributions • Questions • • How do humans and animal think and act ? Contributions • • • • Hermann von Helmholtz : study of human vision (Physiological Optics) H. S. Jennings : Behavior of the lower organisms Wlliam James,… : Cognitive psychology (View of brain as information-processing device) MIT workshop (G. Miller, N. Chomsky, A. Newell and H. Simon) : Cognitive Science (study of mind and its process) (Computer modeling of psychology of memory, language and logical thinking) . . . Computer engineering contributions • Questions • • How can we build an efficient computer ? Contributions • • • • Alan Turing team : First operation computer (Deciphering of German messages) Konarad Zuse : First operation programmable computer, Z-3 John Atanasoff : First electronic computer (ENIAC) Joseph Marie Jacquard : First programmable machine (Loom) using punch cards . . . Control theory and Cybernetics contributions • Questions • • How can artifacts operate under their own control ? Contributions • • • Ktesibios of Alexandia : First self-controlling machine James Watt : Steam engine governor Wiener, Warren, Pitts, John von Neumann : Cybernetics (Scientific study of control and communication in the animal and the machine) • Modern Control Theory : stochastic optimal control that maximizes the objective function over time . . . History of AI • • • • • • • • • Conception of AI Birth of AI Early enthusiasm and expectations Problems in reality Knowledge based Systems AI as Industry Neural networks AI as Science Intelligent agents Conception of AI • Recognition of Warren and Walter’s work as first work in of AI (proposition of artificial neurons) • • • • • • Knowledge of basic physicology Function of neurons in brain Formal analysis of propositional logic Update of connection strengths between neurons (Hebbian learning) First neural network computer (Minsky, Edmonds) Articulation of vision of AI (Computing Machinery and Intelligence by Alan Turing) Birth of AI • • • McCarthy, Minsky, Claude Shannon organized workshops on automata, neural nets, and study of intelligence No break through, just introduction of concerned people with each other AND renaming of the field as Artificial Intelligence Reasons for evolving AI as new field • Duplication of human abilities • • • • Creativity self-improvement Methodology (branch of Computer Science) Aims to build autonomous machines for complex and changing environments Early enthusiasm and expectations • General Problem Solver (GPS) by Newell and Simon • • Limited human thinking capability A series of programs for checkers with learning capabilities • Demonstration on television in February 1956 • Definition of High-level language LISP by McCarthy • Minsky supervised projects known as microworlds • Rearrangements of the blocks on table by the robotic hand Problems in reality • Prediction concerning the future of AI • Computers as chess champions in next 10 years • • Robot football team versus World champions till 2050 • • • • • Achieved in after 40 years (http://en.wikipedia.org/wiki/Robot_Football) • … No background knowledge about problem domain Intractability of the problems to be solved Fundamental limitations on the basic structures for intelligent behavior Combinatorial explosions Knowledge based Systems • Weak methods for solving AI problems • General but do not scale up for problems • • • Expert System designed with specific knowledge from the human expert concerning the field • Expert systems examples • • • • • Large Difficult Accounting Law Medicine process control, etc First neural network computer (Minsky, Edmonds) AI as Science • Building on existing theories • • Why ? Come up with new theories • Prove the theory on all challenging grounds • • • • • Theatrical Mathematical Applicable Feasible, etc Some existing theories concerning AI • • • Hidden Markov Models (HMM) Data mining Bayesian networks etc Intelligent agents • Mobile robots • • Pioneer Humanoids • • Asimo by Honda Nao • Industrial robots • Marine robots, etc State of Art • • • • • Autonomous planning and scheduling Game Playing Autonomous control Diagnosis Image Processing • • • • • Medical Image processing Computer vision etc Logistics Planning Robotics Language understanding etc Summary • • • Foundation of AI History of AI Stat of Art
