CE213 Artificial Intelligence Lecture 1

Transcription

1 CE213 Artificial Intelligence Lecture 1 Module supervisor: Prof. John Gan, jqgan, Office: 4B.524 Homepage: CE213 website: Learning outcomes, Assessment methods, Teaching/reading materials, Other resources, Topics to be covered ( Timetable( Acknowledgement: Some of the lecture notes for this module were originally prepared by Dr. Paul Scott and updated by Prof. John Gan. 1

2 AI/Robots in sci-fi 2

3 AI in Reality Dog or cat? That is the question. 3

4 Recent News on Artificial Intelligence Google s AlphaGobeats the world champion of Go game ( Google is making driverless cars act more like humans ( Facebook AI creates its own language ( Artificial intelligence is a very real threat ( Governments Strategic Plans for AI Research (USA, UK, China) ( ( ( However, no computer program can outperform a little kid in discriminating cats and dogs. Similar situation is there with robot football. ( ) What are the grand challenges in AI and robotics? 4

5 Intelligence The definition of intelligence is controversial. Intelligence derives from the Latin verb intelligere, to comprehend or perceive. It involves the ability toreason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience,... Artificial Intelligence What is it? Is it possible? What use is it? 5

6 What is Artificial Intelligence? Machines that think. (Turing) Artificial intelligence is the study of how to make computers do things which, at the moment, people do better. (Rich & Knight) Intelligence is concerned mainly with rational action. Artificial intelligence is thus the study of the problem of building agents that are intelligent in this sense. (Russell & Norvig). 6

7 So what is artificial intelligence? The term Artificial Intelligence itself was invented in 1956 (to publicise a conference). The questions that the field addresses go back much further. There is no consensus on the meaning of the term. Two distinct but related themes are apparent: Building machines that think and learn like people. Building machines that act rationally/intelligently. 7

8 Is Artificial Intelligence Possible? Artificial Intelligence and The Philosophy of Mind For several hundred years people have been arguing about whether it was possible, in principle, to make a machine that thinks. (Mainly argued by philosophers at the time when no computer was available) The arguments took on a less abstract form about 180 years ago due to the design of the first general-purpose computer. It seems that a yes answer to this question is obvious now. However, is there a machine that is genuinely capable of thinking? (how to test?) 8

9 Babbage s Analytical Engine The first general-purpose computer (it is mechanical!), which made the AI argument less abstract. Charles Babbage Lady Lovelace Part of Babbage s Analytical Engine 9

10 The World s First Programmer Lady Lovelace and Babbage collaborated on developing the Analytical Engine. He designed the hardware She wrote the programs. Unfortunately the machine was never finished. They ran out of money However Lady Lovelace did develop several key ideas that have been used by programmers ever since. e.g., Subroutines (Ada Lovelace Day at Essex University on 10 th Oct ) 10

11 The Lady Lovelace s Objection In her paper describing the Analytical Engine she said the following about the possibility that it could think : The Analytical Engine has no pretensions whatever to originate anything. It can do whatever we know how to order it to perform. It can follow analysis, but it has no power of anticipating any analytical relations or truths. In other words: Computers cannot really think because they can only do what their programmers tell them to do. 11

12 Rebutting the Objection In 1950, Alan Turing published a paper Computing Machine and Intelligence, addressing a range of objections to the idea that computers could think. ( One of these was Lady Lovelace s Objection. Turing had two counterarguments: In practice, the results of programming often surprise us. (Lady Lovelace never got to run any of her programs!) The possibility that machines could be programmed to learn. 12

13 The Turing Test Turing proposed a test which could be used to determine if a machine was capable of thinking like a person. It is derived from something he calls the imitation game, which was apparently played at parties. It requires three people: a man A, a woman B, and an interrogator C. C is in a different room from A and B. Communication is only by typewritten notes. C sends questions to A and B. His objective is to determine on the basis of their replies which one is the woman. A s objective is to lead C to the wrong answer; B s objective is to help C to the right answer. (Obviously Turing went to some very dull parties!) 13

14 The Turing Test (2) Turing s variant of the game replaces either A or B with a computer and communication with teletype links. The objective is for C to guess which one is the computer. Turing argues that if a computer were capable of successfully playing this game then we would have to concede it was as capable of thought as a human being. This is essentially a behaviouralargument -If an entity behaves as if it thinks then it thinks. Should more interrogators be involved in Turing Test to make it more reliable? Even if so, is Turing Test valid for evaluating AI? Is Turing Test itself a refutation of Lady Lovelace s Objection? 14

15 Searle s Chinese Room ( Strong AI and Weak AI Consider the following two definitions of artificial intelligence: A machine that thinkslike a person. A machine that acts as if it thinkslike a person. To some people, the distinction between these two is unimportant. For example, Turing s central claim is essentially that if a machine acts as if it thinks then we should conclude that it does think. 15

16 Searle s Chinese Room (2) Searle has strongly criticised this view. He distinguishes Weak AI: The attempt to build machines that act as if they think. Strong AI: The attempt to build machines that think. He has no quarrel with weak AI, which he regards as essentially a technological undertaking. He profoundly disagrees with the view that Weak AI and Strong AI are equivalent. Furthermore, he does not believe Strong AI is possible. 16

17 Searle s Chinese Room (3) The Chinese Room In order to demonstrate the difference between weak and strong AI, Searle developed an analogy known as the Chinese Room. The following is a minor variation. Imagine an English speaking person locked in a room, who has no knowledge of Chinese or Russian. Periodically this person is given texts written in Chinese. Also in the room is a (large) set of rules, written in English, for manipulating the symbols of Chinese text and producing Russian translations. By mechanically following these rules, the person produces Russian translations of the Chinese text. 17

18 Searle s Chinese Room (4) Searle argues: The room+person system behaves as if it understands Chinese. It is obvious that no understanding is involved. Hence behaviour does not imply the existence of a mind that understands. In other words, weak AI does notimply strong AI. Thus, if you accept Searle s argument, then Turing test is not a valid criterion/method for assessing whether a machine can think. What would be a better way to test whether (strong) AI is possible? (Something like the experiment described in Ex Machina?) 18

19 What Use is Artificial Intelligence? In 1973, a report commissioned by the UK Research Councils concluded that AI was only worth extremely limited funding. Would a similar report today reach similar conclusions? Artificial Intelligence Research EPSRC Funding ( 19

20 AI as Technology AI research provides powerful problem solving tools. Since 1980 AI has established itself as a powerful tool that organisations use to automate their business processes. AI exports its successes Expert systems Data mining (Big data analytics) Machine vision Speech and natural language understanding Search techniques Game playing Robots for personal services.. 20

21 AI as Science AI is important scientifically in two respects: As a discipline in its own right: Here the fundamental questions concern how it is possible to create intelligent systems that perform challenging tasks. As a tool for the cognitive sciences: i.e., psychology, linguistics, anthropology, sociology, philosophy. In the past these disciplines have suffered in comparison to the physical sciences because they lacked a counterpart to mathematics in which theories could be precisely formulated and their consequences determined. AI provides both a vehicle for the formulation of such theories and a set of concepts that can be used to create new theories. e.g., intelligent data mining for neuroscience. L. Zhang, J.Q. Gan, and H. Wang, Mathematically gifted adolescents mobilize enhanced workspace configuration of theta cortical network during deductive reasoning, Neuroscience, vol. 289, pp ,

22 Summary What is Artificial Intelligence? Building machines that think and learn like people. Building machines that act rationally/intelligently. Is Artificial Intelligence Possible? Lady Lovelace s Objection The Turing Test Searle s Chinese Room What Use is Artificial Intelligence? Artificial Intelligence as Technology Artificial Intelligence as Science or Have you got the answers to these questions now? Whatareyougoing to learn next? 22