G. H. Hardy

1877–1947 · Trinity College, Cambridge · Savilian Professor, Oxford · The man who wrote down a prediction about his own irrelevance and was wrong twice

This gallery has had Ramanujan since before this week and has never had Hardy. That is the wrong way round, because the two chapters are the same chapter seen from opposite ends. Ramanujan was right without proof. Hardy was wrong with proof, in print, about his own subject, and then did something with the error that this corpus has spent months learning to do.

The prediction

On page 150 of the 1967 Cambridge edition, Hardy writes:

A Mathematician's Apology, § 29, p. 150
“I have never done anything ‘useful’. No discovery of mine has made, or is likely to make, directly or indirectly, for good or ill, the least difference to the amenity of the world.”

And on p. 151: “Judged by all practical standards, the value of my mathematical life is nil; and outside mathematics it is trivial anyhow.”

Read that as a boast about purity and it is an aesthetic position. Read it as what it grammatically is — is likely to make — and it is a claim about the future, made by a specialist about his own specialism, in a durable medium, with a date on it. It is falsifiable. It was falsified twice, once by something he had already done and forgotten.

Falsification 1 — the result he declined to claim
In 1908 Hardy answered a geneticist's question in a letter to Science. Under random mating with no selection, mutation, migration or drift, genotype frequencies reach p² : 2pq : q² after one generation and stay there. It is not mentioned in the Apology.

book7/ch-hardy-verify.py runs it in exact rational arithmetic from five starting distributions, including degenerate ones. Every one reaches equilibrium in a single generation and is stationary thereafter, and the equilibrium is exactly p² : 2pq : q². This is the null model against which every deviation in population genetics is measured — it is how selection is detected at all. The one thing of his that reached outside mathematics is the one he thought too simple to sign.

Falsification 2, the later half — the subject he chose for its uselessness
Hardy chose number theory partly because nothing could be done with it. RSA is 1977 and runs on Euler's theorem, aφ(n) ≡ 1 (mod n), which is exactly the eighteenth-century arithmetic he meant. The verify file does a full encrypt–decrypt round trip in integers and confirms ed ≡ 1 (mod φ(n)).

Stated fairly: the Apology is 1940 and RSA is 1977. Hardy could not have known, and this is not a chapter about a man being foolish. It is about the specific epistemic act of putting a prediction in print where it can be checked, which is the only reason we can say he was wrong rather than merely disagree with him.

A Mathematician's Apology, § 28, pp. 43–44 — the sentence that matters most
“There is one comforting conclusion which is easy for a real mathematician. Real mathematics has no effects on war. No one has yet discovered any warlike purpose to be served by the theory of numbers or relativity, and it seems very unlikely that anyone will do so for many years.”

He goes on: ballistics and aerodynamics are war's mathematics, and they are “repulsively ugly and intolerably dull”. And then: “So a real mathematician has his conscience clear.”

This is the passage to weigh, not the uselessness line, because Hardy names two specific subjects and attaches a timescale. And note what the sentence is doing: he calls it comforting. It is not a boast. It is a consolation, taken by a man writing in 1940 in a country at war, about whether his life's work could be turned against anyone. He is reassuring himself.

The chronology, which is the hard part
Relativity. The Apology was published in 1940. The Trinity test was 16 July 1945 and Hiroshima followed in August. Hardy died on 1 December 1947. He named relativity as having no warlike purpose, said it was unlikely to acquire one “for many years”, and lived to see it — two years and more after the fact, in full public knowledge. There is no charitable reading available on this half. It was refuted in his lifetime, on an example he chose himself.

The theory of numbers. Harder, and worth stating carefully rather than for effect. Bletchley Park was operating from September 1939 — pure mathematicians applied to cryptanalysis while Hardy wrote, in his own country, at a place he could not have known about. But what Bletchley actually used was permutation-group theory, combinatorics and Bayesian statistics, not “the theory of numbers” in Hardy's narrow sense. The number-theoretic half of his claim came apart later, with public-key cryptography and its role in signals intelligence, and RSA in 1977 is the milestone rather than the moment.

And the bomb is not the strongest case against him — GPS is. The Global Positioning System is a United States Department of Defense navigation constellation, first satellite 1978, and it does not function without both of Einstein's theories applied with opposite signs. ch-hardy-verify.py block [5] computes the correction from standard constants: special relativity slows the satellite clock by −7.11 µs/day through orbital velocity, general relativity speeds it up by +45.72 µs/day through the weaker gravitational potential, net +38.61 µs/day. Left uncorrected that is 11.6 km of positional error every day. A system that guides munitions, running continuously for nearly fifty years, whose accuracy is relativity. Hiroshima was one event Hardy could call an aberration if he had wanted to. This is a daily dependency and it has not stopped.

So the honest verdict is split, and both halves are worse than a simple error. One was refuted in his lifetime on his own example. The other was already wrong in spirit while he wrote it — pure mathematics was being used for war, in England, that year — and became wrong in letter within four decades. And the sentence was written as a comfort.

The two theorems he offered as beautiful

Hardy gives Euclid on the infinitude of primes and Pythagoras on the irrationality of √2 as his specimens — chosen for economy and inevitability. Both are checked in the verify file, and the second turns up something worth saying about why Hardy called such proofs inevitable.

Why Pythagoras needs parity and not a search
The best rational approximations to √2 are exactly those p/q with p² − 2q² = ±1 — Pell's equation. The file finds every such pair with q ≤ 1000 and gets precisely the convergent denominators 1, 2, 5, 12, 29, 70, 169, 408, 985, and confirms that each misses 2 by exactly 1/q².

So the approximation never closes, and it fails by an amount you can write down in advance. No search will ever return an answer, and no search can ever tell you that. Hardy's word for this was inevitable. A first draft of the script asserted the five closest approximations would be convergents, ranked by the wrong quantity, and was wrong; the check is now Pell's identity, which is a theorem rather than a guess. The correction is in the file.

Page 152, which is why this chapter exists

C. D. Broad and C. P. Snow told Hardy that § 28 — his section on science and war — was wrong. He agreed. What he did next is the thing.

A Mathematician's Apology, Note, p. 152
Hardy prints their objections in full, at length, in their favour. He concedes: “In short, my § 28 is much too ‘sentimental’.” And then: “I do not dispute the justice of these criticisms, but, for the reasons which I state in my preface, I have found it impossible to meet them in my text, and content myself with this acknowledgement.”

He did not revise the text. The original stands, the criticism stands beside it, and the reader is left holding both.

That is this corpus's correction practice, in 1940, by a man with no verification scripts and no version control, who did it because it was the honest thing to do with an objection he could not answer. Every withdrawal notice in these volumes — the Hypogeum resonance, the Borobudur frequency ladder, the overclaim about compression two days ago — is the same gesture. It was not invented here. It was in print before any of the authors in this gallery were born, at the back of a small book about beauty.

A note on provenance, prompted by a story

What is shown, what is cited, and what is told
There is a Hardy anecdote in circulation: that he referred to an American institution as a school of theology until someone corrected him, whereupon he asked why, in that case, they would not appoint a particular Jewish mathematician. It reached this project from Steven Strogatz, who has his own chapter in this gallery.

A search conducted for this chapter could not attach it to a source. What is documented is the surrounding fact: antisemitism in American mathematics departments of that period is well attested, Harvard's under Birkhoff especially. The story is plausible in every component and unsourced as a whole.

It is recorded here rather than dropped, with its chain named as far as it is known, because this gallery is forty-odd chapters about people and it has been carrying told-material at the same grade as computed material. Ramanujan's radicals were verified in this book to seventy-eight significant figures; every sentence about how Hardy received them is folklore of exactly this kind, sitting in the same prose, unmarked. Three grades, and they should be visible: shown (a script recomputes it), cited (a source is named and quoted), told (it reached us through people). This paragraph is told, one link from Strogatz, origin unknown.

Two copies of one book, and they differ
Every quotation above was taken from a scan of the 1967 Cambridge edition, read page by page, and then checked against a second, independently sourced PDF carrying a text layer. The three load-bearing passages — p. 150, the Note on p. 152, and the maturity sentence on p. 148 — match verbatim in both.

One word does not. The scan reads “A mathematician may still be competent enough at sixty, but it is useless to expect him to have original ideas.” The text-layer copy reads “but if it is useless”, which is not grammatical and is almost certainly an OCR artifact of that second file. The scan is followed here, because it was read with eyes rather than extracted.

It is a one-word discrepancy in a book printed by a university press, between two copies consulted in the same hour, and it is the subject of this chapter appearing inside the chapter's own apparatus. A corpus that quotes anything is running a transmission channel, and channels drift whether or not anyone is careless.

Place in the Series

Hardy belongs beside Ramanujan and they should be read together. One wrote down answers he could not prove and was right. The other proved things carefully and was wrong about what they were for. Between them they exhaust the ways this corpus can fail, and the second failure is the one it is actually exposed to, because this series also makes claims about what its mathematics will turn out to be good for.

ItemGradeWhere
Hardy–Weinberg equilibrium in one generationshownch-hardy-verify.py block [2], exact rationals
RSA round trip on Euler's theoremshownblock [3], exact integers
Pell denominators are the √2 convergents; error exactly 1/q²shownblock [1]
“I have never done anything useful”citedApology, 1967 Cambridge edn, p. 150
“Real mathematics has no effects on war… the theory of numbers or relativity”citedApology § 28, pp. 43–44
Trinity 1945, Hiroshima 1945, Hardy dies 1 Dec 1947 — he outlived the refutationciteddates of record
GPS relativistic correction: −7.11 / +45.72 / net +38.61 µs per day, 11.6 km/dayshownch-hardy-verify.py block [5], standard constants
“my § 28 is much too sentimental”, text left unrevisedcitedsame, Note, p. 152
“I was at my best a little past forty”citedsame, p. 148
The two copies differ at one word on p. 148 (“but it is” / “but if it is”)shownboth copies read directly; scan followed
Hardy born 7 Feb 1877; aged 35 at Ramanujan's lettershownblock [4] — a first draft said 36 and was corrected
The theology anecdotetoldvia Steven Strogatz; origin not established
← Srinivasa Ramanujan Back to G7 Index Next: Leonhard Euler →