Computing Machinery and Intelligence

Authors: A. M. Turing
Year: 1950 | Mind
Source: turing-1950-computing-machinery-and-intelligence

What it is

Turing’s 1950 paper is where the question “Can machines think?” was first asked and operationalized. Turing didn’t try to define “thinking” — instead he proposed a behavioral test, the Imitation Game, that bypasses the definitional problem entirely. The paper also proposed the “child machine” learning approach that directly anticipates modern machine learning.

The Imitation Game

The original game

Three participants in separate rooms, communicating only through written text:

A man (A) who tries to confuse the interrogator about his identity
A woman (B) who tries to help the interrogator
An interrogator (C) who must determine which is which

The substitution

Replace A with a machine. Now: can the machine fool the interrogator as effectively as a human man would?

This is the Turing Test: if a machine can pass the imitation game, we should credit it with intelligence in every operationally meaningful sense.

Why this framing

Turing’s move was philosophically deliberate. “Can machines think?” requires defining “think” — which leads into unresolvable philosophical debates about consciousness, souls, and subjective experience. The imitation game sidesteps all of that:

“The original question ‘Can machines think?’ I believe to be too meaningless to deserve discussion. Nevertheless I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.”

The test is behavioral and public: it asks what a machine can do, not what it is.

The nine objections

Turing anticipated objections and addressed them — this part of the paper reads as a survey of every argument against machine intelligence.

The theological objection: God gave souls to humans; machines cannot have souls, therefore cannot think.
Turing: This limits what omnipotence can do. God could give souls to machines if He chose.

The mathematical objection (Gödel): By Gödel’s incompleteness theorem, any formal system has statements it cannot prove. A machine running on formal rules is therefore limited in a way humans are not.
Turing: First, humans may face the same limitations. Second, this shows machines have limits, not that those limits are relevant to the imitation game. A machine might still be intelligent enough to pass the test.

Lady Lovelace’s objection: Ada Lovelace wrote that the Analytical Engine “has no power of originating anything. It can only do what we know how to order it to perform.” Machines can only do what they are programmed to do.
Turing: A machine can surprise its programmer. Rules operating on unpredictable inputs produce unpredictable outputs. Surprise is possible even in a deterministic system.

The consciousness objection: A machine cannot feel joy, grief, or love; without these it cannot truly be intelligent.
Turing: We cannot verify consciousness in other humans either. We assume other people are conscious by behavioral analogy. The same logic should apply to machines that behave intelligently.

The child machine proposal

The most practically important part of the paper:

“Instead of trying to produce a programme to simulate the adult mind, why not rather try to produce one which simulates the child’s? If this were then subjected to an appropriate course of education one would obtain the adult brain.”

Turing proposes:

Build a system that starts with some basic structure but minimal knowledge
Define reward and punishment signals to guide learning
Let the system learn from experience and education

This is almost exactly the description of modern deep learning + RLHF:

Random initialization → training on data → evaluation signals → fine-tuning

Turing even notes that the learning process will produce capabilities the programmer did not explicitly specify — which he views as a feature, not a bug.

Turing’s prediction

In 1950, Turing predicted:

“I believe that in about fifty years’ time it will be possible to programme computers, with a storage capacity of about 10⁹, to make them play the imitation game so well that an average interrogator will not have more than 70 per cent chance of making the right identification after five minutes of questioning.”

Fifty years later (2000): the prediction was controversial. Early chatbots (ELIZA, ALICE) could fool some people in short interactions, but no system robustly passed rigorous tests.

Seventy-five years later (2025): large language models routinely produce text indistinguishable from human writing in most casual conversations. Whether this constitutes “passing the Turing test” depends on test design — a debate Turing arguably anticipated.

Why it matters today

The paper did several things that still resonate:

Framed AI as behavioral — what matters is what a system can do, not its internal mechanism. This is the empiricist stance underlying modern ML evaluation.
Anticipated learning over programming — the child machine insight is now consensus. We don’t program knowledge; we train systems.
Named the hardest problems — consciousness, creativity, originality, emotion. These are still the open questions.
Legitimized the field — by taking the question seriously and giving it a rigorous form, Turing made “artificial intelligence” a scientifically respectable research program.

AI/ML Wiki

Explorer

turing-1950-computing-machinery-and-intelligence