JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
Back to the library
Information / CS 1936

On Computable Numbers, with an Application to the Entscheidungsproblem

Alan M. Turing

Turing imagined the simplest possible machine — a head reading a tape — and used it to define computation itself, then proved there are questions no machine can ever answer.

Choose your version
In depth · the introduction

Before any computer existed, Turing described the simplest machine that could compute anything at all — and used it to prove there are questions no computer will ever answer.

The idea, unpacked

Picture a long paper tape divided into squares, a little head that can read and write one square at a time, and a short rulebook. Each rule says: if you are in such-and-such a state and read such-and-such a symbol, write this, step left or right, and switch to that state. That is the whole machine. Turing's claim was startling: anything that can be calculated by a definite procedure can be done by a machine this simple.

Then came the masterstroke. The rulebook for any machine can itself be written out as symbols — so it can be written on a tape and fed to another machine. Turing built one “universal” machine that reads the description of any other machine and then behaves exactly like it. That is software: one device that becomes a word processor, a calculator, or a game, depending only on the description you load. He imagined it a decade before anyone built one.

Where it came from

In 1935 the 22-year-old Turing, a fellow at King's College, Cambridge, sat in on Max Newman's lectures on the open problems of mathematical logic. One was Hilbert's Entscheidungsproblem — the “decision problem” — which asked for a mechanical procedure that could decide, for any mathematical statement, whether it follows from the axioms. Turing took the word “mechanical” literally: to answer it, he first had to say exactly what a machine could do.

He finished the paper in 1936 — then learned that the American logician Alonzo Church had just reached the same conclusion by a completely different method. Turing's was independent and already at the press; he added an appendix proving the two approaches agreed, and soon left for Princeton to study with Church. The image at the heart of his proof — a human reduced to a tireless clerk following rules — would, within fifteen years, become the actual machine he had only imagined.

Why it mattered

It did two opposite things at once. It drew a sharp line under what machines can do — proving that some perfectly clear questions, like “will this program ever stop?”, have no general mechanical answer, so the dream of fully automating mathematics was impossible. And in the very same stroke it described the universal machine: the proof that one general-purpose device could, in principle, run any program at all. Every computer is a realisation of that single idea; the limits and the powers were discovered together, in the same paper.

An analogy

Imagine a clerk at a desk with an endless strip of paper and a thin rulebook. He doesn't understand the problem; he just reads the square in front of him, looks up the matching rule, writes, shuffles one square along, and repeats. Now notice: the rulebook is just writing too. So you could hand the clerk a description of any other clerk's rulebook and ask him to imitate it — one clerk who can stand in for all of them. The unanswerable question is simply: “Given a clerk and his starting page, will he ever put down his pencil?” No rulebook can always tell. Try the tiny machine below — it stops after 14 steps, but its cousins may run forever.

An interactive Turing machine: a tape of cells each showing 0 or 1, with a read/write head and a state badge above the current cell. A Step slider from 0 to 14 advances the machine one rule at a time; the loaded program writes six 1s on a blank tape and halts after 14 steps.

Where it sits

This is the keystone of the Library's information story. George Boole (1854) had turned logic into algebra; Kurt Gödel (1931) had shown that no formal system can prove every truth about arithmetic — and Turing's machine recast that limit in terms anyone can picture. His universal machine is the blueprint von Neumann's 1945 report turned into real hardware, and the bit it shuffles is what Shannon (1948) learned to measure. Fourteen years on, Turing's own 1950 paper asks the sequel: not what these machines cannot do, but whether they could ever think.

The original document
Original source text

Opening — the computable numbers

A. M. Turing · On Computable Numbers · Proc. London Math. Soc. (2) 42 (1936)
The “computable” numbers may be described briefly as the real numbers whose expressions as a decimal are calculable by finite means.
[ … ]

§1–§2 — Computing machines

We may compare a man in the process of computing a real number to a machine which is only capable of a finite number of conditions q1, q2, …, qR which will be called “m-configurations”.
The machine is supplied with a “tape” (the analogue of paper) running through it, and divided into sections (called “squares”) each capable of bearing a “symbol”.
[ … ]

§6 — The universal computing machine

It is possible to invent a single machine which can be used to compute any computable sequence.
If this machine U is supplied with a tape on the beginning of which is written the S.D of some computing machine M, then U will compute the same sequence as M.
[ … ]

§11 — The Entscheidungsproblem

There can be no general process for determining whether a given formula U of the functional calculus K is provable.
[ … ]
Received 28 May 1936 · Read 12 November 1936