Aug 1999 CO230 : Cognitive Science (Question 2) Suggested Solutions

Cognitive and computer scientists have tried to answer the question of whether machines can be intelligent. Numerous tests were designed to prove or disprove the concept.
(a)	Alan Turing designed a test to determine if machines can think. (i) Describe the test. (ii) Briefly describe the strengths of the test. (iii) Briefly describe the weaknesses of the test. (iv) Do you think that Turing Test is useful in determining if machines are intelligent? Explain.	[4] [2] [2] [3]
	(i) The test is conducted as a game with three participants, two persons and a computer. A person acts as an interrogator and is physically separated from the other participant and the computer. To ensure impartiality, the interrogator and the other two entities communicate via electronic means (e.g through terminals). The interrogator will 'asks' the other two entities questions, and through the answers given, will try to identify the human being. If the interrogator identifies the other two entities incorrectly, then the computer is said to be intelligent enough to fool the interrogator.
	(ii) It uses a known source of intelligence(human!!) as a basis of comparison and therefore injects an element of objectivity. It prevents is from being side tracked on issues like the appropriateness of the internal processes of the machine and of the consciousness of the machine of its response. It eliminates any bias in favor of the living organism over machine intelligence by forcing the interrogator to focus on the contents of the answers.
	(iii) Bias towards purely symbolic tasks and ignoring other aspects of human intelligence(like perceptual skill and manual dexterity). The test does not recognize other forms of intelligence besides human intelligence.
	(iv) The test obviously does not determine if machines are intelligent. In the first place, human beings are not the only intelligent life forms. In fact, if we look at plants, since plants can adapt for survival, maybe plants can be termed as intelligent too! However, the fundamental problem is in the definition of 'intelligence'; there is no commonly accepted concept of intelligence. To say if something is intelligent, we have to agree on what intelligence is in the first place.
(b)	John Searle attempted to disprove strong AI by presenting the Chinese Room Argument. (i) What is strong AI? (ii) Describe the Chinese Room Argument. (iii) Briefly describe a possible flaw in the Argument.	[2] [4] [3]
	(i) According to strong AI, an appropriately programmed computer that generates the correct output given the input is deemed to be working like the human mind.
	(ii) Assuming that the role of a computer is taken by people locked inside a room. These people understand no Chinese but receive input in Chinese. Through the use of a rule-book, the input is manipulated, and the output in Chinese is then returned to the people outside. John Searle concluded that even if the correct output is generated, the programmed computer is not like the brain because the output is derived without the elements of understanding.
	(iii) The Argument uses people to take the role of the computer and shows that the people performs symbol manipulation and obtains the correct output without understanding what they are doing. However, linguists had managed to interpret, to a certain extent, ancient languages without even having initial knowledge of the syntactic rules of the language. Given sufficient input, an intelligent linguist may be able to correlate symbols with others to derive understanding of the language. Making use of artificial intelligence and machine learning, it is conceivable that computers can achieve the level of sophistication in the future.