MIND - cs.furman.edu

1 VOL. LIX. NO. 236.] [October, 1950 MINDA QUARTERLY REVIEWOFPSYCHOLOGY AND PHILOSOPHYI. computing machinery ANDINTELLIGENCEBY A. M. TUBING1. The Imitation PROPOSE to consider the question, ' Can machines think ? 'This should begin with definitions of the meaning of the tTms' machine ' a n d' think '. The definitions might be framed so as toreflect so far as possible the normal use of the words, but thisattitude is dangerous. If the meaning of the words ' machine'and' think ' are to be found by examining how they are commonlyused it is difficult to escape the conclusion that the meaningand the answer to the question, ' Can machines think ?' is to besought in a statistical survey such as a Gallup poll. But this isabsurd. Instead of attempting such a definition I shall replace thequestion by another, which is closely related to it and is expressedin relatively unambiguous new form of the problem can be described in terms ofa game which we call the ' imitation game '.

2 It is played withthree people, a man (A), a woman (B), and an interrogator (C) whomay be of either sex. The interrogator stays in a room apartfrom the other two. The object of the game for the interrogatoris to determine which of the other two is the man and which isthe woman. He knows them by labels X and Y, and at the endof the game he says either ' X is A and Y is B ' or ' X is B and Yis A'. The interrogator is allowed to put questions to A and Bthus:C : Will X please tell me the length of his or her hair ?Now suppose X is actually A, then A must answer. It is A's28 3434 A. M. TURING :object in the game to try and cause C to make the wrong identi-fication. His answer might therefore be' My hair is shingled, and the longest strands are about nineinches long.'In order that tones of voice may not help the interrogatorthe answers should be written, or better still, typewritten. Theideal arrangement is to have a teleprinter communicating betweenthe two rooms.

3 Alternatively the question and answers can berepeated by an intermediary. The object of the game for the thirdplayer (B) is to help the interrogator. The best strategy for heris probably to give truthful answers. She can add such thingsas ' I am the woman, don't listen to him t' to her answers, butit will avail nothing as the man can make mmilar now ask the question, ' What will happen when a machinetakes the part of A in this game ?' Will the interrogator decidewrongly as often when the game is played like this as he doeswhen the game is played between a man and a woman ? Thesequestions replace our original,' Can machines think ? '2. Critique of the New well as asking, ' What is the answer to this new form of thequestion', one may ask, ' Is. this new question a worthy oneto investigate ?' This latter question we investigate withoutfurther ado, thereby cutting short an infinite new problem has the advantage of drawing a fairly sharpline between the physical and the intellectual capacities of a engineer or chemist claims to be able to produce a materialwhich is indistinguishable from the human alrin It is possiblethat at some time this might be done, but even supposing this in-vention available we should feel there was little point in tryingto make a ' machine ' more human by dressing it up insuch artificial flesh.

4 The form in which we have set the problemreflects this fact in the condition which prevents the interrogatorfrom seeing or touching the other competitors, or hearing theirvoices. Some other advantages of the proposed criterion may beshown up by specimen questions and answers. Thus:Q: Please write me a sonnet on the subject of | t he : Count me out on this one. I never could write : Add 34957 to 70764A : (Pause about 30 seconds and then give as answer) : Do you play chess ?A: machinery AND intelligence 435Q : I have K at my Kl, and no other pieces. You have onlyK at K6 and R at Rl. It is your move. What do youplay?A : (After a pause of 15 seconds) R-R8 question and answer method seems to be suitable forintroducing almost any one'of the fields of human endeavour thatwe wish to include. We do not wish to penalise the machinefor its inability to shine in beauty competitions, nor to penalisea man for losing in a race against an aeroplane. The conditionsof our game make these disabilities irrelevant.

5 The ' witnesses 'can brag, if they consider it advisable, as much as they pleaseabout their charms, strength or heroism, but the interrogatorcannot demand practical game may perhaps be criticised on the ground that theodds are weighted too heavily against the machine. If the manwere to try and pretend to be the machine he would clearly makea very poor showing. He would be given away at once by slownessand inaccuracy in arithmetic. May not machines carry out some-thing which ought to be described as thinking but which is verydifferent from what a man does ? This objection is a very strongone, but at least we can say that if, nevertheless, a machine canbe constructed to play the imitation game satisfactorily, we neednot be troubled by this might be urged that when playing the ' imitation game'the best strategy for the machine may possibly be somethingother than imitation of the behaviour of a man. This may be, butI think it is unlikely that there is any great effect of this any case there is no intention to investigate here the theoryof the game, and it will be assumed that the best strategy isto try to provide answers that would naturally be given by a The Machines concerned in the question which we put in 1 will not be quite definiteuntil we have specified what we mean by the word ' machine'.

6 It is natural t h at we should wish to permit every kind of engineeringtechnique to be used :n our machines. We also wish to allow thepossibility than an engineer or team of engineers may constructa machine which works, but whose manner of operation cannotbe satisfactorily described by its constructors because they haveapplied a method which is largely experimental. Finally, wewish to exclude from the machines men born in the usual is difficult to frame the definitions so as to satisfy these threeconditions. One might for instance insist that the team of436 A. M. TUBING :engineers should be all of one sex, but this would not reallybe satisfactory, for it is probably possible to rear a completeindividual from a single cell of the skin (say) of a man. To doso would be a feat of biological technique deserving of the veryhighest praise, but we would not be inclined to regard it'as acase of ' constructing a thinting machine'. This prompts us toabandon the requirement that every kind of technique shpuldbe permitted.

7 We are the more ready to do so uLview of thefact that the present interest in ' thinking machines' hatPbeenaroused by a particular kind of machine, usually called an' electronic computer ' or ' digital computer'. Following thissuggestion we only permit digital computers to take part in restriction appears at first sight to be a very drastic shall attempt to show that it is not so in reality. To do thisnecessitates a short account of the nature and properties of may also be said that this identification of machines withdigital computers, like our criterion for ' thinking', will onlybe unsatisfactory if (contrary to my belief), it turns out thatdigital computers are unable to give a good showing in the are already a number of digital computers in workingorder, and it may be asked,' Why not try the experiment straightaway ? It would be easy to satisfy the conditions of the number of interrogators could be used, and statistics compiledto show how often the right identification was given.

8 ' The shortanswer is that we are not asking whether all digital computerswould do well in the game nor whether the computers at presentavailable would do well, but whether there are imaginable com-puters which would do well. But this is only the short shall see this question in a different light Digital idea behind digital computers may be explained by sayingthat these machines are intended to carry out any operationswhich could be done by a human computer. The human computeris supposed to be following fixed rules; he has no authorityto deviate from them in any detail. We may suppose that theserules are supplied in a book, which is altered whenever he is puton to a new job. He has also an unlimited supply of paper onwhich he does his calculations. He may also do his multiplicationsand additions on a ' desk machine', but this is not we use the above explanation as a definition we shall be inCOMPUTING MACKINBBY AND intelligence 437danger of circularity of argument.

9 We avoid this by givingan outline of the means by which the desired effect is digital computer can usually be regarded as consisting of threeparts:(i) Store.(ii) Executive unit,(iii) store is a store of information, and corresponds to the humancomputer's paper, whether this is the paper on which he does hiscalculations or that on which his book of rules is printed. In sofar as the human computer does calculations in his head a part ofthe store will correspond to bis executive unit is the part which carries out the variousindividual operations involved in a calculation. What theseindividual operations are will vary from machine to fairly lengthy operations can be done such as ' Multiply3640676445 by 7076346687' but in some machines only verysimple ones such as ' Write down 0 ' are have mentioned that the ' book of rules' supplied to thecomputer is replaced in the machine by a part of the store. Itis then called the ' table of instructions '. It is the duty of thecontrol to see that these instructions are obeyed correctly and inthe right order.

10 The control is so constructed that this information in the store is usually broken up into packetsof moderately small size. In one machine, for instance, a packetmight consist of ten decimal digits. Numbers are assigned to theparts of the store in which the various packets of informationare stored, in some systematic manner. A typical instructionmight say ' Add the number stored in position 6809 to that in 4302 andput the result back into the latter storage position'.Needless to say it would not occur in the machine expressedin English. It would more likely be coded in a form such as6809430217. Here 17 says which of various possible operationsis. to be performed on the two numbers. In this case the opera-tion is that described above, viz. ' Add the number..' Itwill be noticed that the instruction takes up 10 digits and soforms one packet of information, very conveniently. The controlwill normally take the instructions to be obeyed in the order ofthe positions in which they are stored, but occasionally an in-struction such as438 A.