Turing Test Paper

Alan Turing's paper "Computing Machinery and Intelligence," published in 1950 in the journal Mind, is a foundational work in the field of Artificial intelligence. It introduced to the general public the concept now known as the Turing test, a method for evaluating a machine's ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. The paper directly addresses the question "Can machines think?" and proposes a more concrete alternative based on observable performance.

Turing argued that the terms "think" and "machine" are too ambiguous for a meaningful debate. To resolve this, he suggested replacing the original question with one expressed in relatively unambiguous words. He outlined a three-step approach: first, substitute a simple concept for "think"; second, specify the types of machines under consideration; and third, pose a new question that he believed could be answered affirmatively.

The Imitation Game

Turing's proposed substitute was a game he called the "Imitation Game," originally a party game with three players: a man (A), a woman (B), and an interrogator (C) who can be of either sex. The interrogator, isolated from the other two, communicates with them only through written notes and must determine which player is the man and which is the woman. Player A tries to deceive the interrogator, while player B assists the interrogator in making the correct identification.

Turing then proposed a variation where a computer takes the role of player A. The question becomes whether the interrogator would decide wrongly as often when a machine is involved as when the game is played between a man and a woman. This modified game involves a computer, a human, and a human judge, all in separate rooms. The judge converses with both via a terminal, and both the computer and the human attempt to convince the judge they are human. If the judge cannot reliably distinguish the computer from the human, the computer wins.

This formulation shifted the focus from whether a machine can "think" in a philosophical sense to whether it can act indistinguishably from a thinking entity. As noted by cognitive scientist Stevan Harnad, the question became "Can machines do what we (as thinking entities) can do?" This approach avoids the difficulty of pre-defining "think" and instead emphasizes performance capacities. Since its introduction, the test has been highly influential and widely criticized, becoming a central concept in the philosophy of artificial intelligence. Criticisms, such as John Searle's Chinese room argument, are themselves subjects of debate. Turing's intent was not merely to fool a person but to generate human cognitive capacity.

Digital Machines

Turing also specified which machines were to be considered. He dismissed human clones as uninteresting examples and focused on digital computers - machines that manipulate binary digits (0 and 1) using simple rules. He gave two reasons for this choice. First, digital computers already existed in 1950, so there was no need to speculate about their feasibility. Second, digital machinery is "universal": Turing's earlier work on computation had shown that a digital computer can, in theory, simulate the behavior of any other digital machine, given enough memory and time. This is the essence of the Church-Turing thesis and the universal Turing machine. Consequently, if any digital machine could act as if thinking, then every sufficiently powerful digital machine could. Turing wrote, "all digital computers are in a sense equivalent."

He restated the question more specifically: "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 emphasized that the focus was not on whether all or currently available computers would pass the test, but on whether imaginable computers could, considering potential future advancements.

Nine Common Objections

Turing addressed nine common objections to the possibility of machine thinking, which encompassed major arguments raised against Artificial intelligence in subsequent years:

1. Religious Objection: Thinking is a function of the immortal soul, so machines cannot think. Turing responded that constructing machines is not irreverently usurping divine power, just as procreation is not; rather, both are instruments of divine will.

1. 'Heads in the Sand' Objection: The consequences of machines thinking would be too dreadful. Turing noted this fear is common among intellectuals who see superiority in intelligence and fear being overtaken.

1. Mathematical Objection: Limitations of formal systems, such as Gödel's incompleteness theorems, suggest machines cannot match human mathematical insight. Turing argued that humans also make mistakes and that these limitations apply to any formal system, including human reasoning.

1. Argument from Consciousness: A machine must be conscious and have emotions to truly think. Turing replied that this is solipsistic, as we can only verify consciousness in ourselves, and the test focuses on external behavior.

1. Arguments from Various Disabilities: Machines cannot do certain things like be kind, resourceful, or fall in love. Turing countered that these are not well-defined and that machines might eventually exhibit such behaviors.

1. Lady Lovelace's Objection: The Analytical Engine can only do what we tell it, so it cannot originate anything. Turing argued that machines can surprise us, as they can produce results beyond what programmers explicitly intended.

1. Argument from Continuity in the Nervous System: The brain is not a discrete-state machine, so digital computers cannot replicate it. Turing acknowledged differences but noted that discrete machines can simulate continuous systems closely enough.

1. Argument from Informality of Behaviour: Human behavior is too informal and unpredictable to be captured by rules. Turing suggested that this informality might be simulated through learning, not pre-programmed rules.

1. Argument from Extrasensory Perception: If ESP exists, it could complicate the test. Turing dismissed this as not relevant to the main question.

Legacy and Impact

The paper laid the groundwork for the field of Artificial intelligence and the Turing test became a benchmark for machine intelligence. It influenced later developments in Machine learning, Neural network research, and the creation of Large language models. Turing's ideas about digital machines and their universality also connected to the broader theory of computation. The paper remains a touchstone in discussions of AI ethics and capability, and its questions continue to shape research in Generative AI and beyond.