Artificial Intelligence

Transcription

1 Artificial Intelligence (Sistemas Inteligentes) Pedro Cabalar Depto. Computación Universidade da Coruña, SPAIN Chapter 1. Introduction Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 1 / 13

2 Outline 1 A definition of AI Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 2 / 13

3 1 A definition of AI Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 3 / 13

4 A bit of debate: what is AI? Take 5 min. to tell what Artificial Intelligence (AI) is all about... Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 4 / 13

5 A bit of debate: what is AI? Take 5 min. to tell what Artificial Intelligence (AI) is all about... Debate: half class will be defenders and half class attackers. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 4 / 13

6 A bit of debate: what is AI? Take 5 min. to tell what Artificial Intelligence (AI) is all about... Debate: half class will be defenders and half class attackers. What is AI? It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable. John McCarthy node1.html Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 4 / 13

7 A bit of debate: what is AI? What do you think? Give arguments why AI is a... 1 science 2 engineering 3 a constant disappointment Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 5 / 13

8 What is AI? A definition of AI A keypoint: what is intelligence? Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 6 / 13

9 What is AI? A definition of AI A keypoint: what is intelligence? A definition will depend on human intelligence and we ignore many of its mechanisms. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 6 / 13

10 What is AI? A definition of AI A keypoint: what is intelligence? A definition will depend on human intelligence and we ignore many of its mechanisms. The perception of intelligent behavior has changed along History. Example: a calculator looked intelligent a hundred years ago while some wouldn t say a chess program looks intelligent today. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 6 / 13

11 What is AI? A definition of AI A keypoint: what is intelligence? A definition will depend on human intelligence and we ignore many of its mechanisms. The perception of intelligent behavior has changed along History. Example: a calculator looked intelligent a hundred years ago while some wouldn t say a chess program looks intelligent today. Many definitions of AI have been made. They can be classified as follows: Imitating human behavior: thinking like humans vs acting like humans Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 6 / 13

12 What is AI? A definition of AI A keypoint: what is intelligence? A definition will depend on human intelligence and we ignore many of its mechanisms. The perception of intelligent behavior has changed along History. Example: a calculator looked intelligent a hundred years ago while some wouldn t say a chess program looks intelligent today. Many definitions of AI have been made. They can be classified as follows: Imitating human behavior: thinking like humans vs acting like humans Focus on rational behaviour: thinking rationally vs acting rationally Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 6 / 13

13 Acting humanly: the Turing test Alan Turing ( ) The Turing test: we have two terminals A=controlled by a computer; B=with a human behind. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 7 / 13

14 Acting humanly: the Turing test Alan Turing ( ) The Turing test: we have two terminals A=controlled by a computer; B=with a human behind. C is a human interrogator that must find out who is who. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 7 / 13

15 Acting humanly: the Turing test Alan Turing ( ) The Turing test: we have two terminals A=controlled by a computer; B=with a human behind. C is a human interrogator that must find out who is who. We say A is intelligent = C cannot tell. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 7 / 13

16 Acting humanly: the Turing test Can you imagine what A should be capable of? Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 8 / 13

17 Acting humanly: the Turing test Can you imagine what A should be capable of? Natural language Knowledge representation Automated reasoning Machine learning Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 8 / 13

18 Acting humanly: the Turing test Can you imagine what A should be capable of? Natural language Knowledge representation Automated reasoning Machine learning Total Turing test: includes video signal, perception and exchange of physical objects... Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 8 / 13

19 Acting humanly: the Turing test Can you imagine what A should be capable of? Natural language Knowledge representation Automated reasoning Machine learning Total Turing test: includes video signal, perception and exchange of physical objects... Computer Vision Robotics Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 8 / 13

20 Acting humanly: the Turing test Can you imagine what A should be capable of? Natural language Knowledge representation Automated reasoning Machine learning Total Turing test: includes video signal, perception and exchange of physical objects... Computer Vision Robotics These are the six main areas of AI and became the real goal, rather than the test itself. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 8 / 13

21 Thinking humanly: Cognitive modeling Cognitive Science tries to join AI models with experimental techniques from Psychology to build (testable) theories about the human mind. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 9 / 13

22 Thinking humanly: Cognitive modeling Cognitive Science tries to join AI models with experimental techniques from Psychology to build (testable) theories about the human mind. Two ways of tackling the problem of cognitive modeling: 1 Symbolic modeling: Use knowledge-based systems to capture abstract mental functions handling symbols. Marvin Minsky s school. Marvin Minsky (1927-) Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 9 / 13

23 Thinking humanly: Cognitive modeling Cognitive Science tries to join AI models with experimental techniques from Psychology to build (testable) theories about the human mind. Two ways of tackling the problem of cognitive modeling: 1 Symbolic modeling: Use knowledge-based systems to capture abstract mental functions handling symbols. Marvin Minsky s school. Marvin Minsky (1927-) 2 Subsymbolic modeling: try to follow the neural and associative properties of the human brain using connectionst/neural network models. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 9 / 13

24 Thinking humanly: Cognitive modeling Cognitive Science tries to join AI models with experimental techniques from Psychology to build (testable) theories about the human mind. Two ways of tackling the problem of cognitive modeling: 1 Symbolic modeling: Use knowledge-based systems to capture abstract mental functions handling symbols. Marvin Minsky s school. Marvin Minsky (1927-) 2 Subsymbolic modeling: try to follow the neural and associative properties of the human brain using connectionst/neural network models. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade da Chapter Coruña, 1. SPAIN Introduction ) 9 / 13

25 Thinking rationally: laws of thought Logicist tradition in AI. John McCarthy s school. John McCarthy ( ) Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 10 / 13

26 Thinking rationally: laws of thought Logicist tradition in AI. John McCarthy s school. John McCarthy ( ) Logic: solid background since Aristotle. Three chronological eras: Philosophy, Mathematics and Computational Logic. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 10 / 13

27 Thinking rationally: laws of thought Logicist tradition in AI. John McCarthy s school. John McCarthy ( ) Logic: solid background since Aristotle. Three chronological eras: Philosophy, Mathematics and Computational Logic. Obstacles: too rigid for dealing with uncertainty; high computational cost for practical problems. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 10 / 13

28 Thinking rationally: laws of thought Logicist tradition in AI. John McCarthy s school. John McCarthy ( ) Logic: solid background since Aristotle. Three chronological eras: Philosophy, Mathematics and Computational Logic. Obstacles: too rigid for dealing with uncertainty; high computational cost for practical problems. All AI systems must face these same obstacles, but they appeared first in the logicist tradition. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 10 / 13

29 Acting rationally: rational agent Agent = something that acts Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 11 / 13

30 Acting rationally: rational agent Agent = something that acts Computer agents are expected to: operate autonomously, perceiving the environment, adapting to change, taking on another s goals, etc. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 11 / 13

31 Acting rationally: rational agent Agent = something that acts Computer agents are expected to: operate autonomously, perceiving the environment, adapting to change, taking on another s goals, etc. A rational agent should achieve the best outcome or, when there is uncertainty, the best expected outcome. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 11 / 13

32 Acting rationally: rational agent Agent = something that acts Computer agents are expected to: operate autonomously, perceiving the environment, adapting to change, taking on another s goals, etc. A rational agent should achieve the best outcome or, when there is uncertainty, the best expected outcome. Note that making correct inferences (logicist approach) is sometimes part of a rational agent. Other actions (example: reflect reactions) can also be rational but not inferential. Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 11 / 13

33 Acting rationally: rational agent Agent = something that acts Computer agents are expected to: operate autonomously, perceiving the environment, adapting to change, taking on another s goals, etc. A rational agent should achieve the best outcome or, when there is uncertainty, the best expected outcome. Note that making correct inferences (logicist approach) is sometimes part of a rational agent. Other actions (example: reflect reactions) can also be rational but not inferential. Computational limitations make perfect rationality unachievable: design best program for given machine resources Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 11 / 13

34 AI prehistory Philosophy Mathematics Psychology Economics Linguistics Neuroscience Control theory A definition of AI logic, methods of reasoning mind as physical system foundations of learning, language, rationality formal representation and proof algorithms, computation, (un)decidability, (in)tractability probability adaptation phenomena of perception and motor control experimental techniques (psychophysics, etc.) formal theory of rational decisions knowledge representation grammar plastic physical substrate for mental activity homeostatic systems, stability simple optimal agent designs Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 12 / 13

35 Potted history of AI 1943 McCulloch & Pitts: Boolean circuit model of brain 1950 Turing s Computing Machinery and Intelligence Look, Ma, no hands! 1950s Early AI programs, including Samuel s checkers program, Newell & Simon s Logic Theorist, Gelernter s Geometry Engin 1956 Dartmouth meeting: Artificial Intelligence adopted 1965 Robinson s complete algorithm for logical reasoning AI discovers computational complexity Neural network research almost disappears Early development of knowledge-based systems Expert systems industry booms Expert systems industry busts: AI Winter Neural networks return to popularity 1988 Resurgence of probability; general increase in technical dept Nouvelle AI : ALife, GAs, soft computing 1995 Agents, agents, everywhere Human-level AI back on the agenda Pedro Cabalar (UDC) ( Depto. AIComputación Universidade dachapter Coruña, 1. SPAIN Introduction ) 13 / 13