Computing machinery and intelligence

1. The Imitation Game: A New Lens on Machine Thought

I PROPOSE to consider the question, ‘Can machines think?’

Turing's Proposal. Instead of directly tackling the ambiguous question of whether machines can think, Turing introduces the "imitation game" as a practical test. This game involves a human interrogator trying to distinguish between a human and a machine based on their written responses. The core idea is that if a machine can consistently fool the interrogator, it demonstrates a form of intelligence comparable to human thought.

Game Mechanics. The imitation game involves three participants: a man (A), a woman (B), and an interrogator (C). The interrogator's goal is to identify which of the other two is the man and which is the woman by asking questions. The challenge is then posed: can a machine take the place of the man and still deceive the interrogator as often as a human could?

Shifting the Focus. By framing the problem in terms of a game, Turing sidesteps the need for precise definitions of "machine" and "think." The focus shifts to observable behavior and the ability to mimic human responses, providing a concrete and measurable criterion for assessing machine intelligence. This approach allows for a more pragmatic and less philosophical exploration of the question.

2. Distinguishing Physical from Intellectual Prowess

The new problem has the advantage of drawing a fairly sharp line between the physical and the intellectual capacities of a man.

Intellectual vs. Physical. Turing emphasizes that the imitation game effectively separates the intellectual abilities from the physical attributes of a human. The game's design prevents the interrogator from relying on physical cues like appearance, voice, or touch, forcing them to focus solely on the content and quality of the responses. This distinction is crucial for evaluating whether a machine can truly "think" rather than simply mimic human-like physical characteristics.

Irrelevant Disabilities. The conditions of the imitation game render certain human disabilities irrelevant. A machine is not penalized for its inability to win beauty contests, nor is a man penalized for losing a race against an airplane. The game focuses on intellectual performance, allowing for a fair comparison between humans and machines in a specific domain.

Bragging Rights. The game allows participants to "brag" about their charms, strength, or heroism, but the interrogator cannot demand practical demonstrations. This highlights the importance of persuasive communication and the ability to create a convincing narrative, further emphasizing the intellectual aspects of the test.

3. Defining the "Machine" in Thinking Machines

The question which we put in § 1 will not be quite definite until we have specified what we mean by the word ‘machine’.

Engineering Techniques. Turing acknowledges the need to define "machine" to clarify the scope of the inquiry. He intends to permit every kind of engineering technique to be used in the machines, including those whose manner of operation cannot be satisfactorily described by their constructors due to experimental methods. However, he excludes men born in the usual manner.

Digital Computers. Turing narrows the definition of "machine" to digital computers, recognizing that the current interest in thinking machines has been aroused by this particular type of machine. This restriction appears drastic at first, but Turing argues that it is not so in reality, necessitating a short account of the nature and properties of these computers.

The Short Answer. Turing addresses the question of why not try the experiment straight away with existing digital computers. He clarifies that the inquiry is not about whether all or the currently available digital computers would do well in the game, but whether there are imaginable computers which would do well, setting the stage for a more theoretical exploration.

4. Digital Computers as Universal Simulators

The idea behind digital computers may be explained by saying that these machines are intended to carry out any operations which could be done by a human computer.

Mimicking Human Computers. Turing explains that digital computers are designed to perform any operation that a human computer could carry out, following fixed rules without deviation. These rules are supplied in a book, which is altered whenever the computer is put on a new job. This analogy helps to illustrate the fundamental concept of a digital computer as a machine capable of executing instructions.

Three Parts. A digital computer consists of three parts: a store, an executive unit, and a control. The store holds information, corresponding to the human computer's paper and memory. The executive unit carries out individual operations, and the control ensures that instructions are obeyed correctly and in the right order.

Instruction Tables. The "book of rules" is replaced by a "table of instructions" stored in the machine. The control unit follows these instructions, which are coded in a form that specifies the operation to be performed and the storage locations involved. This allows for complex sequences of operations to be repeated until a condition is fulfilled, mimicking the actions of a human computer.

5. The Universality Property and Its Implications

This special property of digital computers, that they can mimic any discrete state machine, is described by saying that they are universal machines.

Discrete State Machines. Turing classifies digital computers as "discrete state machines," which move by sudden jumps from one definite state to another. Although strictly speaking there are no such machines, many can profitably be thought of as such. This abstraction allows for a simplified analysis of their behavior.

Mimicking Behavior. Given the table corresponding to a discrete state machine, it is possible to predict what it will do. This calculation can be carried out by a digital computer, which can mimic the behavior of any discrete state machine if it has adequate storage capacity and works sufficiently fast.

Universal Machines. This special property of digital computers, that they can mimic any discrete state machine, is described by saying that they are universal machines. This has the important consequence that it is unnecessary to design various new machines to do various computing processes, as they can all be done with one digital computer, suitably programmed for each case.

6. Addressing Objections to Machine Intelligence

We may now consider the ground to have been cleared and we are ready to proceed to the debate on our question, ‘Can machines think?’

Contrary Views. Turing acknowledges that opinions will differ as to the appropriateness of replacing the original question with the imitation game. He proceeds to consider various objections to the idea of machine intelligence, addressing each in turn.

Theological Objection. The theological objection posits that thinking is a function of man's immortal soul, which God has not given to animals or machines. Turing counters this by arguing that it implies a restriction of the Almighty's omnipotence, suggesting that God could confer a soul on an elephant or a machine if He saw fit.

Heads in the Sand Objection. The "Heads in the Sand" objection expresses the fear that the consequences of machines thinking would be too dreadful. Turing dismisses this argument as insufficiently substantial, suggesting that consolation should be sought in the transmigration of souls.

7. The Theological and Consciousness Arguments

Thinking is a function of man’s immortal soul.

Mathematical Objection. The mathematical objection uses results from mathematical logic, such as Gödel's theorem, to show limitations to the powers of discrete-state machines. Turing argues that while there are limitations to any particular machine, it has not been proven that no such limitations apply to the human intellect.

Argument from Consciousness. The argument from consciousness, expressed by Professor Jefferson, asserts that a machine cannot truly think until it can write a sonnet or compose a concerto because of thoughts and emotions felt. Turing counters this by suggesting that it leads to solipsism, where one can only be sure that oneself thinks.

Viva Voce. Turing uses the example of a viva voce examination to illustrate how the imitation game can be used to discover whether someone truly understands something or has merely learned it parrot fashion. He argues that if a machine can answer questions in a satisfactory and sustained manner, it would be difficult to dismiss it as merely artificially signaling.

8. Learning Machines: Mimicking the Child's Mind

Instead of trying to produce a programme to simulate the adult mind, why not rather try to produce one which simulates the child’s?

Lady Lovelace's Objection. Turing revisits Lady Lovelace's objection, which stated that a machine can only do what we tell it to do. He counters this by suggesting that a man can "inject" an idea into the machine, which will respond to a certain extent and then drop into quiescence. However, he also proposes the possibility of a "super-critical" machine that can generate a whole theory from an injected idea.

Skin of an Onion Analogy. Turing introduces the "skin of an onion" analogy, where we find certain operations of the mind that can be explained in purely mechanical terms. Stripping off these layers, we may eventually come to the "real" mind or find that the whole mind is mechanical.

Child-Programme. Turing suggests that instead of trying to produce a program to simulate the adult mind, it may be easier to produce one that simulates the child's mind. This child-program could then be subjected to an appropriate course of education to obtain the adult brain.

9. The Role of Programming and Education

As I have explained, the problem is mainly one of programming.

Programming Advances. Turing emphasizes that the problem of creating thinking machines is mainly one of programming. Advances in engineering will also be necessary, but it seems unlikely that these will not be adequate for the requirements.

Storage Capacity. Estimates of the storage capacity of the brain vary from 10^10 to 10^15 binary digits. Turing inclines to the lower values and believes that only a very small fraction is used for the higher types of thinking. He suggests that a storage capacity of 10^9 would be required for satisfactory playing of the imitation game.

Child-Programme and Education. Turing divides the problem into two parts: the child-program and the education process. He suggests that one must experiment with teaching one such machine and see how well it learns, then try another and see if it is better or worse.

10. The Significance of Randomness and Exploration

It is probably wise to include a random element in a learning machine.

Evolutionary Connection. Turing draws a connection between the process of teaching a child-machine and evolution, with the structure of the child machine corresponding to hereditary material, changes corresponding to mutations, and the judgment of the experimenter corresponding to natural selection.

Punishments and Rewards. Turing associates punishments and rewards with the teaching process. He suggests that some simple child-machines can be constructed or programmed on this sort of principle, where events shortly preceding a punishment-signal are unlikely to be repeated, whereas a reward-signal increased the probability of repetition.

Random Element. Turing suggests that it is probably wise to include a random element in a learning machine. A random element is useful when searching for a solution to some problem, and the learning process can be regarded as a search for a form of behavior that will satisfy the teacher.

Last updated: April 8, 2025