The Imitation Game and Lady Lovelace's Objection
Why Playing Chess, Learning From Mistakes, and Conversing Replaced Definitions of Thought
Asking whether a machine can think seems like a natural starting point for the philosophy of mind. Yet when Alan Turing opened his landmark paper Computing Machinery and Intelligence in the October 1950 issue of the philosophical review Mind, he dismissed the question as too ambiguous to merit scientific debate. Attempting to define the words machine and think by polling common linguistic usage reduces a technical investigation into an empty dictionary dispute. Instead, Turing replaced metaphysical definitions with a concrete, falsifiable behavioral experiment that he called the Imitation Game.
In its baseline setup, the game involves three participants isolated in separate rooms: a man $A$, a woman $B$, and a human interrogator $C$ of either sex. To prevent vocal pitch or physical appearance from giving away the answer, all communication takes place strictly through typewritten teleprinter messages passed between the rooms. The interrogator $C$ asks unconstrained questions—ranging from everyday habits to poetry composition and arithmetic—while $A$ tries to deceive the interrogator and $B$ tries to help the interrogator identify who is who. Turing then posed his operational replacement question. What happens when a programmed digital computer takes the place of player $A$ in the game? Specifically, how often will the interrogator guess wrongly compared to when a human is playing?
Because a digital computer is a physical realization of his 1936 Universal Turing Machine, Turing proved that constructing a specialized mechanical brain is unnecessary. Any discrete-state machine can be mimicked to arbitrary fidelity by programming a single universal computer equipped with sufficient memory capacity. He estimated that roughly $10^{9}$ to $10^{10}$ binary digits of storage would suffice to fool an average interrogator across a five-minute teleprinter interrogation.
In the centerpiece of the 1950 paper, Turing anticipated and dismantled nine major philosophical, mathematical, and biological objections to machine intelligence. Four technical rebuttals formed the core of his argument:
- * The Mathematical Objection: Critics invoke Kurt Gödel's 1931 incompleteness theorem and Turing's own 1936 halting problem to argue that any fixed formal machine must fail on certain self-referential questions. Turing replied that human mathematicians frequently make calculation errors of their own. Moreover, humans possess zero proven immunity to simultaneous incompleteness limits across the human brain.
- * The Argument from Consciousness: Philosophers such as Geoffrey Jefferson argued in his 1949 Lister Oration that no machine truly thinks until it writes a sonnet out of felt emotion. Turing pointed out that taking that standard literally leads straight to solipsism. After all, the only way any person verifies consciousness in another human being is by observing external conversation and behavior.
- * Lady Lovelace's Objection: In her 1843 memoir on Charles Babbage's Analytical Engine, Augusta Ada, Countess of Lovelace, stated that the machine has no pretensions to originate anything. It can only execute whatever we know how to order it to perform. Turing answered that deterministic rule tables routinely generate consequences that surprise their human designers. Programmers simply cannot compute millions of intermediate state steps in their heads in advance.
- * The Argument from Fallibility and Randomness: Critics object that a strict rule-following computer can never make a creative slip or vary its responses. Turing noted that a programmer can easily insert a pseudo-random digit generator or allow learning rules to modify their own transition tables through experience.
In the closing pages of the paper, building on his 1948 NPL report Intelligent Machinery, Turing calculated that hand-coding a complete adult mind would be far too brittle. Instead of attempting to write a static program for an adult chess player or conversationalist from scratch, the final pages of Turing's 1950 paper outlined the training of an unorganized child machine through reward, punishment, and empirical experience.