Transcript PPT
CSCI-100 Introduction to Computing Artificial Intelligence What is AI? • Artificial intelligence (AI) is usually defined as the science of making computers do things that require intelligence when done by humans What is Intelligence? • The ability to think and act rationally • The capacity to learn – Consider the behavior of the digger wasp. When the wasp brings food to her nest, she puts it on the entrance, goes inside to check for intruders and, if the coast is clear, carries in the food. If you move the food a few inches while the wasp is inside checking, on emerging, the wasp repeats the whole procedure (i.e., it carries the food to the entrance, goes in to look around, and emerges again) Dumb insect What’s involved in Intelligence? • Ability to interact with the real world – to perceive, understand, and act – speech recognition, understanding – Image understanding (computer vision) • Reasoning and planning – Modeling the external world – Problem solving, planning, and decision making – Ability to deal with unexpected problems, uncertainties • Learning and Adaptation What’s involved in Intelligence? • Research in AI has focused mainly on the following components of intelligence – – – – – Learning Reasoning Problem Solving Perception Language Understanding Strong AI • Strong AI aims to build machines whose overall intellectual ability is indistinguishable from that of a human being – The ultimate goal of [strong] AI is nothing less than to build a machine on the model of a man, a robot that is to have its childhood, to learn a language as a child does, to gain its knowledge of the world by sensing the world through its own organs, and ultimately to contemplate the whole domain of human thought [ Joseph Weizenbaum, MIT AI Laboratory ] Applied AI • Applied AI aims to produce commercially viable “smart” systems (e.g., a security system that is able to recognize the faces of people who are permitted to enter a particular building) • Applied AI has enjoyed considerable success Different Approaches Against human performance Thinking Behavior Against ideal concept of intelligence (rationality) Systems that think like humans Systems that think rationally Systems that act like humans Systems that act rationally Thinking Humanly: Cognitive Science • Claim: A given program thinks like a human – Must have some way of determining how humans think – Need to get inside the actual workings of human minds – After we have a theory of the mind, can express it as a computer program – If program’s I/O and timing matches corresponding human behavior, evidence that our theory is correct Systems that think like humans Systems that think rationally Systems that act like humans Systems that act rationally Thinking Humanly: Cognitive Science • The interdisciplinary field of cognitive science brings together computer models from AI and experimental techniques from psychology to try to construct precise and testable theories of the workings of the human mind Thinking Rationally: Laws of Thought • Aristotle – Attempted to codify “right thinking” (i.e., correct arguments) – His syllogisms provided patterns for argument structures that always yielded correct conclusions given correct premises Socrates is a man All men are mortal Socrates is mortal • By 1965, programs existed that could, in principle, solve any solvable problem described in logical notation • The logicist tradition within AI hopes to build on such programs to create intelligent systems Systems that think like humans Systems that think rationally Systems that act like humans Systems that act rationally Acting Humanly: The Turing Test ELIZA, computer therapist Systems that think like humans Systems that think rationally Systems that act like humans Systems that act rationally Acting Humanly: The Turing Test • Proposed by Alan Turing (1950) (the father of AI) • Based on indistinguishability from undeniably intelligent entities (i.e., we) • Is a computer that passes the test really intelligent? • Programming a computer to pass the test provides plenty to work on – – – – Natural Language Processing Knowledge Representation Automated Reasoning Machine Learning Acting Humanly: The Turing Test • Physical simulation of a person unnecessary for intelligence thus no need for physical interaction between interrogator and computer • Total Turing Test – Interrogator can test the subject’s perceptual abilities – Interrogator can pass physical objects “through the hatch” – Computer would need computer vision, robotics Acting Humanly: The Turing Test • Today, AI researchers devote little effort to passing the test • More important to study underlying principles of intelligence • Turing Test suggested major components of AI – – – – – – Natural Language Processing Knowledge Representation Automated Reasoning Machine Learning Computer Vision Robotics Acting Rationally: Rational Agents • 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 Systems that think like humans Systems that act like humans Systems that think rationally Systems that act rationally Acting Rationally: Rational Agents • An agent is an entity that perceives and acts • Examples of Agents? • Sensors and Actuators in… – A Human Agent? – A Robotic Agent? Acting Rationally: Rational Agents (Advantages) • In the “Laws of Thought” approach to AI, emphasis is on correct inferences • Making correct inferences is sometimes part of being a rational agent but not all of rationality. Why? – Sometimes there is no provably correct thing to do, yet something must still be done – There are ways of acting rationally that cannot be said to involve inference (e.g., recoiling from a hot stove) • Thus, more general than the “Laws of Thought” approach Acting Rationally: Rational Agents (Advantages) • More suitable for scientific development that approaches based on human behavior or human thought because standard of rationality is clearly defined and completely general AI History • History of AI is commonly supposed to begin with Turing’s 1950 discussions of machine intelligence and to have been defined as a field at the 1956 Dartmouth Summer Research Project on Artificial Intelligence – But ideas on which AI is based, symbolic AI in particular, have a very long history in the Western intellectual tradition, dating back to ancient Greece Symbolic AI • Approach to AI that has dominated the field throughout most of its history • Based on the Physical Symbol System Hypothesis, enunciated by Newell and Simon (1976) – “A physical symbol system has the necessary and sufficient means for general intelligent action” – Knowledge represented in brain by language-like structures or formulas – Thinking is a computational process that rearranges such structures according to formal rules The Roots of Formal Logic • In ancient Greece, pebbles were used for calculation in a similar way to beads on an abacus – Latin word for “pebble” is calculus – In logic and math, we use the word “calculus” for any system of notation in which we can accomplish some purpose by manipulation of tokens according to formal, mechanical rules (e.g., propositional, predicate calculus) – To the extent that such rules are purely mechanical, they can, in principle, be carried out by a machine • If a process can be reduced to a calculus, it can be calculated by a machine Neuroscience • Neuroscience (1861 – present) – How do brains process information? • • • • • • Brain consists of brain nerve cells or neurons A neuron makes connections with other neurons at junctions called synapses Signals are propagated from neuron to neuron The signals enable long-term changes in connectivity of neurons Thought to form the basis for learning in the brain A collection of simple cells can lead to thought, action, and consciousness or, in other words, that brains cause minds (Searle, 1992) • Alternative is mysticism: there is a mystical realm in which minds operate that is beyond physical science Connectionism • Has only recently become a serious contender to symbolic AI • The level of the symbol is too high to lead to a good model of the mind – Have to go lower: instead of designing programs that perform computations on such symbols, design programs that perform computations at a lower level (the neuron) – When viewed at the semantic levels such systems often do not appear to be engaged in rule-following behavior (rules lie at a deeper level)