PROPOSAL FOR RESEARCH BY M.L.MINSKY
It isnot difficult to design a machine which exhibits the following type oflearning. The machine is provided with input and output channels and aninternal means of providing varied output responses to inputs in such a waythat the machine may be ``trained'' by a ``trial and error'' process to acquireone of a range of input-output functions. Such a machine, when placed in anappropriate environment and given a criterior of ``success'' or ``failure'' canbe trained to exhibit ``goal-seeking'' behavior. Unless the machine is providedwith, or is able to develop, a way of abstracting sensory material, it canprogress through a complicated environment only through painfully slow steps,and in general will not reach a high level of behavior.
Now letthe criterion of success be not merely the appearance of a desired activitypattern at the output channel of the machine, but rather the performance of agiven manipulation in a given environment. Then in certain ways the motorsituation appears to be a dual of the sensory situation, and progress can bereasonably fast only if the machine is equally capable of assembling anensemble of ``motor abstractions'' relating its output activity to changes in theenvironment. Such ``motor abstractions'' can be valuable only if they relate tochanges in the environment which can be detected by the machine as changes inthe sensory situation, i.e., if they are related, through the structure of theenvironrnent, to the sensory abstractions that the machine is using.
I havebeen studying such systems for some time and feel that if a machine can bedesigned in which the sensory and motor abstractions, as they are formed, canbe made to satisfy certain relations, a high order of behavior may result.These relations involve pairing, motor abstractions with sensory abstractionsin such a way as to produce new sensory situations representing the changes inthe environment that might be expected if the corresponding motor act actuallytook place.
Theimportant result that would be looked for would be that the machine would tendto build up within itself an abstract model of the environment in which it isplaced. If it were given a problem, it could first explore solutions within theinternal abstract model of the environment and then attempt externalexperiments. Because of this preliminary internal study, these externalexperiments would appear to be rather clever, and the behavior would have to beregarded as rather ``imaginative''
A verytentative proposal of how this might be done is described in my dissertationand I intend to do further work in this direction. I hope that by summer 1956 Iwi11 have a model of such a machine fairly close to the stage of programming ina computer.
PROPOSAL FOR RESEARCH BY N. ROCHESTER
Originality in Machine Performance
Inwriting a program for an automatic calculator, one ordinarily provides themachine with a set of rules to cover each contingency which may arise andconfront the machine. One expects the machine to follow this set of rulesslavishly and to exhibit no originality or common sense. Furthermore one isannoyed only at himself when the machine gets confused because the rules he hasprovided for the machine are slightly contradictory. Finally, in writingprograms for machines, one sometimes must go at problems in a very laboriousmanner whereas, if the machine had just a little intuition or could makereasonable guesses, the solution of the problem could be quite direct. Thispaper describes a conjecture as to how to make a machine behave in a somewhatmore sophisticated manner in the general area suggested above. The paperdiscusses a problem on which I have been working sporadically for about fiveyears and which I wish to pursue further in the ArtificialIntelligence Project next summer.
The Process of Invention or Discovery
Livingin the environment of our culture provides us with procedures for solving manyproblems. Just how these procedures work is not yet clear but I shall discussthis aspect of the problem in terms of a model suggested by Craik . He suggests that mental action consists basically ofconstructing little engines inside the brain which can simulate and thuspredict abstractions relating to environment. Thus the solution of a problemwhich one already understands is done as follows:
1. Theenvironment provides data from which certain abstractions are formed.
2. Theabstractions together with certain internal habits or drives provide:
2.1 Adefinition of a problem in terms of desired condition to be achieved in thefuture, a goal.
2.2 Asuggested action to solve the problem.
2.3 Stimulationto arouse in the brain the engine which corresponds to this situation.
3. Thenthe engine operates to predict what this environmental situation and theproposed reaction will lead to.
4. Ifthe prediction corresponds to the goal the individual proceeds to act asindicated.
Theprediction will correspond to the goal if living in the environment of hisculture has provided the individual with the solution to the problem. Regardingthe individual as a stored program calculator, the program contains rules tocover this particular contingency.
For amore complex situation the rules might be more complicated. The rules mightcall for testing each of a set of possible actions to determine which providedthe solution. A still more complex set of rules might provide for uncertainty aboutthe environment, as for example in playing tic tac toe one must not onlyconsider his next move but the various possible moves of the environment (hisopponent).
Nowconsider a problem for which no individual in the culture has a solution andwhich has resisted efforts at solution. This might be a typical currentunsolved scientific problem. The individual might try to solve it and find thatevery reasonable action led to failure. In other words the stored programcontains rules for the solution of this problem but the rules are slightlywrong.
Inorder to solve this problem the individual will have to do something which isunreasonable or unexpected as judged by the heritage of wisdom accumulated bythe culture. He could get such behavior by trying different things at randombut such an approach would usually be too inefficient. There are usually toomany possible courses of action of which only a tiny fraction are acceptable.The individual needs a hunch, something unexpected but not altogether reasonable.Some problems, often those which are fairly new and have not resisted mucheffort, need just a little randomness. Others, often those which have longresisted solution, need a really bizarre deviation from traditional methods. Aproblem whose solution requires originality could yield to a method of solutionwhich involved randomness.
Interms of Craik's S model, the engine which should simulate the environment atfirst fails to simulate correctly. Therefore, it is necessary to try variousmodifications of the engine until one is found that makes it do what is needed.
Insteadof describing the problem in terms of an individual in his culture it couldhave been described in terms of the learning of an immature individual. Whenthe individual is presented with a problem outside the scope of his experiencehe must surmount it in a similar manner.
So farthe nearest practical approach using this method in machine solution ofproblems is an extension of the Monte Carlo method. In the usual problem which isappropriate for Monte Carlo there is a situation which is grossly misunderstoodand which has too many possible factors and one is unable to decide whichfactors to ignore in working out analytical solution. So the mathematician hasthe machine making a few thousand random experiments. The results of theseexperiments provide a rough guess as to what the answer may be. The extensionof the Monte Carlo Method is to use these results as a guide to determine whatto neglect in order to simplify the problem enough to obtain an approximateanalytical solution.
Itmight be asked why the method should include randomness. Why shouldn't themethod be to try each possibility in the order of the probability that thepresent state of knowledge would predict for its success? For the scientistsurrounded by the environment provided by his culture, it may be that onescientist alone would be unlikely to solve the problem in his life so theefforts of many are needed. If they use randomness they could all work at onceon it without complete duplication of effort. If they used system they wouldrequire impossibly detailed communication. For the individual maturing incompetition with other individuals the requirements of mixed strategy (usinggame theory terminology) favor randomness. For the machine, randomness willprobably be needed to overcome the shortsightedness and prejudices of theprogrammer. While the necessity for randomness has clearly not been proven,there is much evidence in its favor.
TheMachine With Randomness
Inorder to write a program to make an automatic calculator use originality itwill not do to introduce randomness without using forsight. If, for example,one wrote a program so that once in every 10,000 steps the calculator generateda random number and executed it as an instruction the result would probably bechaos. Then after a certain amount of chaos the machine would probably trysomething forbidden or execute a stop instruction and the experiment would beover.
Twoapproaches, however, appear to be reasonable. One of these is to find how thebrain manages to do this sort of thing and copy it. The other is to take someclass of real problems which require originality in their solution and attemptto find a way to write a program to solve them on an automatic calculator.Either of these approaches would probably eventually succeed. However, it isnot clear which would be quicker nor how many years or generations it wouldtake. Most of my effort along these lines has so far been on the former approachbecause I felt that it would be best to master all relevant scientificknowledge in order to work on such a hard problem, and I already was quiteaware of the current state of calculators and the art of programming them.
Thecontrol mechanism of the brain is clearly very different from the controlmechanism in today's calculators. One symptom of the difference is the mannerof failure. A failure of a calculator characteristically produces somethingquite unreasonable. An error in memory or in data transmission is as likely tobe in the most significant digit as in the least. An error in control can donearly anything. It might execute the wrong instruction or operate a wronginput-output unit. On the other hand human errors in speech are apt to resultin statements which almost make sense (consider someone who is almost asleep,slightly drunk, or slightly feverish). Perhaps the mechanism of the brain issuch that a slight error in reasoning introduces randomness in just the rightway. Perhaps the mechanism that controls serial order in behavior guides the random factor so as to improve the efficiency ofimaginative processes over pure randomness.
Somework has been done on simulating neuron nets on our automatic calculator. Onepurpose was to see if it would be thereby possible to introduce randomness inan appropriate fashion. It seems to have turned out that there are too manyunknown links between the activity of neurons and problem solving for thisapproach to work quite yet. The results have cast some light on the behavior ofnets and neurons, but have not yielded a way to solve problems requiringoriginality.
Animportant aspect of this work has been an effort to make the machine form andmanipulate concepts, abstractions, generalizations, and names. An attempt wasmade to test a theory3 of how the brain does it. The first set ofexperiments occasioned a revision of certain details of the theory. The secondset of experiments is now in progress. By next summer this work will befinished and a final report will have been written.
Myprogram is to try next to write a program to solve problems which are membersof some limited class of problems that require originality in their solution.It is too early to predict just what stage I will be in next summer, or just;how I will then define the immediate problem. However, the underlying problemwhich is described in this paper is what I intend to pursue. In a singlesentence the problem is: how can I make a machine which will exhibitoriginality in its solution of problems?
REFERENCES
1.K.J.W. Craik, The Nature of Explanation, Cambridge University Press,1943 (reprinted 1952), p. 92.
2. K.S.Lashley, ``The Problem of Serial Order in Behavior'', in Cerebral Mechanismin Behavior, the Hixon Symposium, edited by L.A. Jeffress, John Wiley &Sons, New York, pp. 112-146, 1951.
3. D.O. Hebb, The Organization of Behavior, John Wiley & Sons, New York,1949
PROPOSAL FOR RESEARCH BY JOHN MCCARTHY
Duringnext year and during the Summer Research Project on Artificial Intelligence, Ipropose to study the relation of language tointelligence. It seems clear that the direct application of trialand error methods to the relation between sensory data and motor activity willnot lead to any very complicated behavior. Rather it is necessary for the trialand error methods to be applied at a higher level of abstraction. The humanmind apparently uses language as its means of handling complicated phenomena.The trial and error processes at a higher level frequently take the form offormulating conjectures and testing them. The English language has a number ofproperties which every formal language described so far lacks.
1. Arguments in English supplemented byinformal mathematics can be concise.
2. English is universal in the sense thatit can set up any other language within English and then use that languagewhere it is appropriate.
3. The user of English can refer tohimself in it and formulate statements regarding his progress in solving theproblem he is working on.
4. In addition to rules of proof, Englishif completely formulated would have rules of conjecture .
Thelogical languages so far formulated have either been instruction lists to makecomputers carry out calculations specified in advance or else formalization ofparts of mathematics. The latter have been constructed so as:
1. to be easily described in informalmathematics,
2. to allow translation of statements frominformal mathematics into the language,
3. to make it easy to argue about whetherproofs of (???)
Noattempt has been made to make proofs in artificial languages as short asinformal proofs. It therefore seems to be desirable to attempt to construct anartificial language which a computer can be programmed to use on problemsrequiring conjecture and self-reference. It should correspond to English in thesense that short English statements about the given subject matter should haveshort correspondents in the language and so should short arguments orconjectural arguments. I hope to try to formulate a language having theseproperties and in addition to contain the notions of physical object, event,etc., with the hope that using this language it will be possible to program amachine to learn to play games well and do other tasks.
PEOPLE INTERESTED IN THE ARTIFICIAL INTELLIGENCE PROBLEM
Thepurpose of the list is to let those on it know who is interested in receivingdocuments on the problem. The people on the 1ist wlll receive copies of thereport of the Dartmouth Summer Project on Artificial Intelligence. [1996 note:There was no report.]
Thelist consists of people who particlpated in or visited the Dartmouth SummerResearch Project on Artificlal Intelligence, or who are known to be interestedin the subject. It is being sent to the people on the 1ist and to a few others.
For thepresent purpose the artificial intelligence problem is taken to be that ofmaking a machine behave in ways that would be called intelligent if a humanwere so behaving.
Arevised list will be issued soon, so that anyone else interested in getting onthe list or anyone who wishes to change his address on it should write to:
John McCarthy
Dapartment of Mathematics
Dartmouth College
Hanover, NH
[1996note: Not all of these people came to the Dartmouth conference. They werepeople we thought might be interested in Artificial Intelligence.] (Mr. Qinlongji notes 47 p.)
The list consists of:
Adelson,Marvin
HughesAircraft Company, Airport Station, Los Angeles, CA
Ashby, W.R.
BarnwoodHouse, Gloucester, England
Backus,John
IBMCorporation, 590 Madison Avenue, New York, NY
Bernstein,Alex
IBMCorporation, 590 Madison Avenue, New York, NY
Bigelow,J. H.
Institutefor Advanced Studies, Princeton, NJ
Elias,Peter
R. L.E., MIT, Cambridge, MA
Duda, W.L.
IBMResearch Laboratory, Poughkeepsie, NY
Davies,Paul M.
1317 C. 18thStreet, Los Angeles, CA.
Fano, R.M.
R. L.E., MIT, Cambridge, MA
Farley,B. G.
324 ParkAvenue, Arlington, MA.
Galanter,E. H.
Universityof Pennsylvania, Philadelphia, PA
Gelernter,Herbert
IBMResearch, Poughkeepsie, NY
Glashow,Harvey A.
1102Olivia Street, Ann Arbor, MI.
Goertzal,Herbert
330 West11th Street, New York, New York
Hagelbarger,D.
BellTelephone Laboratories, Murray Hill, NJ
Miller,George A.
MemorialHall, Harvard University, Cambridge, MA.
Harmon,Leon D.
BellTelephone Laboratories, Murray Hill, NJ
Holland,John H.
E. R. I., University of Michigan
AnnArbor, MI
Holt,Anatol, 7358 Rural Lane, Philadelphia, PA
Kautz,William H.
StanfordResearch Institute, Menlo Park, CA
Luce, R.D.
427 West117th Street, New York, NY
MacKay,Donald
Departmentof Physics, University of London, London, WC2, England
McCarthy,John
DartmouthCollege, Hanover, NH
McCulloch,Warren S.
R.L.E., M.I.T., Cambridge, MA
Melzak,Z. A.
MathematicsDepartment, University of Michigan
AnnArbor, MI
Minsky,M. L. , 112 Newbury Street, Boston, MA
More,Trenchard
Departmentof Electrical Engineering, MIT, Cambridge, MA
Nash, John
Institutefor Advanced Studies, Princeton, NJ
Newell,Allen
Departmentof Industrial Administration, Carnegie Institute of Technology, Pittsburgh, PA
Robinson,Abraham
Departmentof Mathematics, University of Toronto, Toronto, Ontario, Canada
Rochester,Nathaniel
EngineeringResearch Laboratory, IBM Corporation, Poughkeepsie, NY
Rogers,Hartley, Jr.
Departmentof Mathematics, MIT, Cambridge, MA.
Rosenblith,Walter
R.L.E.,M.I.T. , Cambridge, MA.
Rothstein,Jerome
21 EastBergen Place, Red Bank, NJ
Sayre,David
IBMCorporation, 590 Madison Avenue, New York, NY
Schorr-Kon,J.J.
C-380Lincoln Laboratory, MIT, Lexington, MA
Shapley,L.
RandCorporation, 1700 Main Street, Santa Monica, CA
Schutzenberger,M.P.
R.L.E.,M.I.T. , Cambridge, MA
Selfridge,O. G.
LincolnLaboratory, M.I.T. , Lexington, MA
Shannon,C. E.
R.L.E.,M.I.T. , Cambridge, MA
Shapiro,Norman
RandCorporation, 1700 Main Street, Santa Monica, CA
Simon,Herbert A.
Departmentof Industrial Administration, Carnegie Institute of Technology, Pittsburgh, PA
Solomonoff,Raymond J.
TechnicalResearch Group, 17 Union Square West, New York, NY
Steele,J. E., Capt. USAF
Area B.,Box 8698, Wright-Patterson AFB, Ohio
Webster,Frederick
62Coolidge Avenue, Cambridge, MA
Moore, E.F.
BellTelephone Laboratory, Murray Hill, NJ
Kemeny,John G.
DartmouthCollege, Hanover, NH
[Mr. Qin notes: 13 pages original paper (PDF).]
Aboutthis document ...
Next:Aboutthis document
John McCarthy
Wed Apr 3 19:48:31 PST 1996 |
微信扫一扫
