00 · The Map

This chapter is the courtroom itself — the two benches, the rules of examination, and the standing of the witnesses. Every chapter after it is a hearing.

Why are philosophy and computing the same project, separated at birth?

For three days in September 1930, the limits of reason were argued twice in the same town by people who did not quite hear each other. The Second Conference on the Epistemology of the Exact Sciences met in Königsberg from 5 to 7 September; John von Neumann came to present the formalist case for securing mathematics by finite mechanical means (conference proceedings, Erkenntnis 2, 1931). On the last day, in a general discussion, Kurt Gödel let slip — off-hand, almost in passing — that he could exhibit arithmetical propositions that were true and unprovable in the formal system of classical mathematics. Von Neumann understood at once what had just been said (SEP, Gödel's Incompleteness Theorems). The next day, 8 September, a few streets away, David Hilbert closed an address to the Society of German Natural Scientists and Physicians with six words later cut into his tombstone: Wir müssen wissen. Wir werden wissen. We must know. We shall know.

The remark and the vow were about the same thing, and the remark won. Hilbert's program was not arithmetic housekeeping; it was a philosophical wager — that every mathematical question is decidable, that certainty can be made mechanical (SEP, Hilbert's Program). Gödel published in 1931 and the wager was lost. Five years later the rest of the answer arrived, and it arrived in the shape of a machine.

That project was already two and a half centuries old. Gottfried Wilhelm Leibniz spent his life on a device and a dream: a characteristica universalis, a notation in which every concept decomposes into primitives, joined to a calculus ratiocinator, a rule-system for computing with them. The point was not tidiness but peace. "The only way to rectify our reasonings," he wrote, "is to make them as tangible as those of the Mathematicians, so that we can find our error at a glance, and when there are disputes among persons, we can simply say: Let us calculate, without further ado, to see who is right" (Wiener, Leibniz: Selections, 1951, p. 51). The Latin is one word: calculemus. He also built the hardware: a hand-cranked calculator that multiplied and divided, one specimen of which surfaced in a Göttingen attic in 1879.

The line from that dream to the machine on your desk is not a clean one. George Boole did not learn that Leibniz had anticipated parts of his algebra until more than a year after The Laws of Thought (1854), when Robert Leslie Ellis pointed it out. Frege, by contrast, invoked the calculus ratiocinator by name and claimed the inheritance deliberately (SEP, Leibniz's Influence on 19th Century Logic). The dream was rediscovered at least as often as it was handed down — itself evidence for this book's claim: two disciplines keep arriving at the same idea because they are working the same seam.

The seam runs like this. Frege's Begriffsschrift (Halle, 1879) supplied quantifiers, and with them a logic strong enough to carry mathematical reasoning rather than merely classify syllogisms. Russell's letter of 16 June 1902 showed that Frege's foundation contained a contradiction. Hilbert and Ackermann's Grundzüge der theoretischen Logik (1928) posed the Entscheidungsproblem: is there a procedure that decides, for any statement of first-order logic, whether it is valid? Gödel answered a neighbouring question in 1931 with a no. Then in 1936 the Entscheidungsproblem itself fell twice, independently: Alonzo Church in April (American Journal of Mathematics 58: 345–363), and Alan Turing in a paper received 28 May.

Turing's method is the hinge of this whole book. To prove that no procedure exists, he first had to say what a procedure is, and "mechanical" had never been formally defined. So he defined it: an idealized clerk with a paper tape, a finite set of states, and a table of rules, and argued that this exhausts what any human computer following fixed instructions could do. The universal machine, and therefore the stored-program computer, is a by-product of that argument. Nobody was trying to invent an industry. Someone was trying to settle whether reason has a decision procedure, and the computer fell out of the proof.

The estrangement came fast, and it was institutional rather than intellectual. Turing published the founding paper of computing in a mathematics journal, then in October 1950 put the founding question of machine minds in Mind, a philosophy journal — one man, two faculties, fourteen years apart. By the time the first Department of Computer Sciences in the United States was created at Purdue in October 1962, it was established inside the Division of Mathematical Sciences: the new field took the machine from mathematics and left the question with philosophy (Rice & Rosen, Purdue CS history). Departments are cheap to found and expensive to bridge.

Each side then forgot something the other was holding. Computing forgot meaning. Claude Shannon opened information theory in 1948 by setting semantics aside — "These semantic aspects of communication are irrelevant to the engineering problem" — an honest bracket, exactly right for building telephone networks. But a bracket held for eighty years hardens into a metaphysics, and a field that measures information without meaning eventually builds systems whose behavior it can quantify and cannot interpret. Philosophy forgot feasibility. Its standard epistemic agent knows every consequence of what it knows — logical omniscience, a creature no resource bound permits. Whole arguments about knowledge, rationality, and belief were built on an agent that could not be implemented at any scale (Aaronson, Why Philosophers Should Care About Computational Complexity, 2011).

So this book is not a survey with two halves. It is a cross-examination, and the difference matters. A survey reports what each side believes; a cross-examination makes each side answer in the other's vocabulary, where its habits give it no protection. Three rules hold here. The examiner sets the terms — a witness allowed to answer in its own words is being interviewed, not examined. Both voices appear in every chapter: where the machine is on the stand it answers back, and where philosophy is on the stand it objects. And every claim carries its evidence — a date, a work, a source you can open — because the fastest way to lose a cross-examination is to assert what the record will not support.

Part I, chapters 01–08, puts the machine on the stand and lets philosophy ask: what kind of thing is a program, does a machine's answer count as knowledge, how does code mean, can a decision be delegated without losing its morality. Part II, chapters 09–15, reverses the benches: computation asks philosophy what it can actually deliver, and the answers often embarrass positions that sounded secure. No verdict is handed down at the end. Chapter 15 explains why the examination cannot close, and chapter 99 hands you the route to keep running it yourself.

How to read: seventeen files, each self-contained, principles rather than coverage. Every chapter opens with one line placing it in the two courts and closes with Connections — which chapters cross-examine it — plus open questions and a short, checkable reading list. Read straight through, or enter anywhere; the Connections lines are the corridors between courtrooms.

Connections

Everything connects here and this chapter stands on nothing. Chapter 01 is the closest neighbour — logic is where the two disciplines were still one family, and where Gödel, Turing, and Curry–Howard get handled properly. Chapter 09 returns to 1936 to ask what the Church–Turing thesis actually is: mathematics, empirical claim, or definition. Chapter 10 turns philosophy's forgotten feasibility into a standing objection. Chapter 11 audits Shannon's bracket in full.

Open questions

Was the estrangement necessary, or merely administrative? A discipline that had kept its philosophers might have reached interpretability research decades earlier — or might have argued instead of building. Is the courtroom the right shape at all, or does an adversarial framing flatter both sides by assuming they are separable? And if the two are one project, why has no institution managed to house them together for long?

Go deeper

• Martin Davis, The Universal Computer: The Road from Leibniz to Turing, W. W. Norton, 2000 (paperback retitled Engines of Logic, 2001) — this chapter's arc at book length, told through the people.
• Alan M. Turing, "On Computable Numbers, with an Application to the Entscheidungsproblem," Proceedings of the London Mathematical Society, ser. 2, 42 (1936–37): 230–265 — doi.org/10.1112/plms/s2-42.1.230. Read alongside "Computing Machinery and Intelligence," Mind LIX, no. 236 (October 1950): 433–460 — doi.org/10.1093/mind/LIX.236.433 — for the split running through a single mind.
• The Philosophy of Computer Science, Stanford Encyclopedia of Philosophy — the standing map of the borderland; pair it with the entries on the Church–Turing thesis and Hilbert's program.