Alan Turing — "We are not interested in the fact that the machine can do well, but in the fact …"
We are not interested in the fact that the machine can do well, but in the fact that it can do badly.
We are not interested in the fact that the machine can do well, but in the fact that it can do badly.
Click any product to generate a realistic preview. Up to 3 at a time.
* Initial load can take up to 90 seconds — revising the preview in another color is nearly instant.
"The future of computing lies in the development of intelligent machines."
"A man provided with paper, pencil, and rubber, and subject to strict discipline, is in effect a universal machine."
"It is not possible to produce a machine which will be intelligent in the same way that a human being is intelligent."
"The machine should be able to understand what it is doing."
"It is possible for a machine to have a memory in the sense that a human being has a memory."
Found in 1 providers: grok
1 source checked
A machine that only succeeds proves it was built correctly — that's engineering, not intelligence. What's revealing is how a machine fails: does it err in recognizable, human-like ways, or in cold mechanical ones? Errors expose the underlying logic, assumptions, and gaps in a system. Genuine intelligence is identifiable not by its wins but by the nature and pattern of its mistakes.
Turing's 1950 paper proposed the imitation game specifically to evaluate intelligence through failure — a machine convincingly mimicking human mistakes was his benchmark for thinking. At Bletchley Park, he exploited Enigma operator errors, not the cipher's successes, to crack it open. His framework for artificial intelligence was built on the insight that where a system breaks down tells you more about it than where it performs as designed.
In the early 1950s, computers were room-sized calculators celebrated for deterministic accuracy — their value was eliminating human error. The Cold War demanded reliable, error-free military computation. Claiming that a machine's failures were more interesting than its successes was genuinely subversive. AI as a discipline barely existed; cognition was considered exclusively biological. Turing's focus on machine fallibility anticipated probabilistic AI, neural networks, and decades of research into how systems fail in revealing ways.
AI-generated insights based on extensive research and information for context. Factual errors? Email [email protected].
Your cart is empty