Machine and Thought: The Turing Test Instructor: Viola Schiaffonati April, 7 th 2016
Machines and thought 2 The dream of intelligent machines The philosophical-scientific tradition The official birth of Artificial Intelligence (1956) The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle so precisely described that a machine can be made to simulate it (McCarthy 1955)
To start 3 Turing test as starting point Computing Machinery and Intelligence (1950) The imitation game and possible objections Some references to the following debate Even if they are not all explicitly discussed in Turing s paper, they derive from it
To continue 4 Two possible interpretations for the Turing test Original Imitation Game Test, Standard Turing test Consequences in evaluating intelligence Open issues
Can machines think? 5 Problem of definition Machine Thought No single answer Question substituted with another question, connected to the first one and not ambiguous New form of the problem described as a game
The imitation game 6 Three players A man (A), a woman (B), and an interrogator (C) C stays in a room apart from A and B that C knows only as X and Y C has to determine which of the other two is the man and which is the woman just by means of questions A s object: to try and cause C to make the wrong identification B s object: to help the interrogator
The Turing test 7 Can machines think? replaced by What will happen when a machine takes the part of A in this game? Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman?
Why a new question? 8 Is this new question a worthy one to investigate? Precise criterion for success Sharp line between physical and intellectual capacities Question and answer method suitable for introducing almost any one of the fields of human endeavor that we wish to include
Machines concerned in the game 9 Whether there are imaginable computers which would do well in the game Digital computers Able to carry out any operation which could be done by a human computer The human computer is supposed to be following fixed rules Discrete-state machines Store, executive unit, control
Considering again the question 10 Can machines think? replaced by Are there imaginable digital computers which would do well in the imitation game? Let us fix our attention on one particular digital computer C. Is it true that by modifying this computer to have an adequate storage, suitably increasing its speed of action, and providing it with an appropriate programme, C can be made to play satisfactorily the part of A in the imitation game, the part of B being taken by a man?
Turing s answer 11 I believe that in about fifty years time it will be possible to programme computers [ ] to make them play the imitation game so well that an average interrogator will not have more that 70 per cent chance of making the right identification after five minutes of questioning Importance of conjectures for suggesting useful lines of research
Some objections 12 Arguments from various disabilities Machines can make all, except X Turing: skepticism deriving from the idea of machines able to do only repetitive tasks Lady Lovelace s objection Machines can never do anything really new or surprising Turing: machines take me by surprise because I do not do sufficient calculation to decide what to expect them to do Mathematical objection Limits to the capacities of state-discrete machines (Gödel 1931, Turing 1937) Turing: same limits could hold for human reasoning as well
Learning machines 13 How to program machines able to do well in the imitation game? Evolutionary learning Punishments and rewards method Which are the best fields to start with? Abstract or concrete activities? Both approaches
Turing s two tests for intelligence 14 Sterrett (2000) Original Imitation Game test Standard Turing Test Turing presented them as equivalent, Sterrett claims they are not equivalent It is the first, neglected, test that provides a more appropriate indication of intelligence
The two tests (Sterrett 2000) ORIGINAL IMITATION GAME TEST 15 STANDARD TURING TEST What will happen when a machine takes the part of A in this game? Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman? Are there imaginable digital computers which would do well in the imitation game? Is it true that [ ] can one build a particular computer to play satisfactorily the part of A in the imitation game, the part of B being taken by a man?
Original imitation game test 16 Test structure permits the result that the machine does better than the man There is nothing inherent in the game to prevent the machine from scoring higher than the man Test tends to screen off lack of interrogator skill Since the machine s intelligence is measured by comparing the frequency with which it succeeds in causing the interrogator to make the wrong identification with the frequency with which a man does so, the results will not be too sensitive to the skill of the interrogator Both man and machine are required to impersonate. The machine s performance is not directly compared to the man s, but their rates of successfully impersonating against a real woman candidate are compared
Standard Turing test 17 No meaningful result could indicate that the machine does better than the man What test result would indicate that the machine had outperformed the human, give that the criterion is simply giving a performance indistinguishable from a human s? Test results are very sensitive to the interrogator s skills or lack of skill The machine s fortune in passing the test will go up and down with the skill level of the interrogator Only the computer is attempting to impersonate. The computer s performance is judged based on similarity to a man s performance
Intellectual abilities and cognitive habits 18 Better characterization of intelligence with the Original Imitation Game Test To fool the interrogator C, better cognitive abilities are required A must impersonate a woman and induce to make C the wrong identification (both in the case A is a man and in the case A is a machine) This requires the ability to evaluate the adequacy of own answers Impersonation requires higher intellectual abilities
Some advantages? 19 To diminish the man s advantage in the game The test focuses on a notion of machine intelligence, rather than similarity to human Critical cognitive capacities are emphasized, whereas habitual behaviors are de-emphasized Difference between a response requiring thought and a response that, though entirely appropriate, is habitual To de-emphasize training and emphasize thinking
Consequences 20 Intelligence intended in a more general meaning, not merely as human intelligence It is not required to have lived a human like life The original imitation Game Test extracts a supercritical kind of thought from humans, and then uses it to construct a measure by which the machine s capabilities can be measured
References 21 Sterrett, S. (2000) Turing's Two Tests for Intelligence, Minds and Machines 10, 541-59. Turing, A. (1950) Computing machinery and intelligence, Mind 59, 433-460.