How information became the hidden logic of everything

1

Redundancy beats brevity

Talking-drum languages proved that piling on stock phrases, not sharpening precision, is what keeps a signal intact across noise and distance.

Nineteenth-century accounts, especially John Carrington's studies of Kele drum languages, show drummers reproducing only the tonal contours of speech. A single drummed word for chicken could match dozens of unrelated words because drums lack consonants and vowels. Reliability came from wrapping every message in formulaic identifiers of village, family, and cause, far more material than the spoken equivalent. Gleick treats this as the book's founding principle: information systems are judged by how reliably they deliver meaning despite interference, not by how little they say. The same tension between compression and protection reappears in every later chapter, from telegraph codes to genetic error correction. The Congo basin solution anticipated Shannon's formal proof without any mathematics. Modern instincts still prize brevity; the drums reveal that what looks like waste is often the load-bearing structure that lets a message survive.

In practice: A death announcement was drummed with repeated stock phrases naming the village, family, and cause so the full message stayed clear even though any single beat pattern stayed ambiguous.

2

Words become countable objects

Writing and print froze shifting speech into fixed units that could be disputed, stored, and measured, a precondition for later quantitative theories of language.

Before writing, meaning lived only in speakers' memory and shifted with context. Written forms, then print, let people point at a word and argue over it. Samuel Johnson's 1755 dictionary attempted one man's catalog of English; the Oxford English Dictionary, begun in 1857 under James Murray and finished in 1928, went further by collecting thousands of citation slips showing actual historical use and building definitions from evidence rather than decree. That method turned the dictionary into an information system for storing and retrieving a language's usage history. Once words were stable, countable objects, questions became askable: how many words exist, how often each appears, how much uncertainty remains after the previous letter. Those are precisely the questions Shannon later answered mathematically. Print culture's deeper gift was not literacy alone but the idea that language could be studied like any physical phenomenon instead of merely lived inside.

In practice: OED editors gathered citation slips from readers across centuries and derived definitions from observed use rather than prescribing ideal meanings.

3

Messages leave their bodies

The telegraph proved information could cross a continent in seconds while every physical object stayed put, separating fact from transport for the first time.

Samuel Morse developed his code between 1837 and 1844; the first long-distance message traveled from Washington to Baltimore in May 1844. Before that moment a message moved only as fast as horse, ship, or train. Afterward a fact could outrun its own paper. Commerce, journalism, and warfare reorganized almost immediately around near-simultaneous stock prices, war news, and shipping schedules. Because companies charged by the word, operators and customers invented compressed phrasing and shared code books that substituted one code-word for whole common phrases—an informal compression driven by price, not theory. By 1866 a permanent transatlantic cable linked Europe and North America, and information itself was bought and sold by the letter. Like the drummers, telegraphers worked constantly against cost and noise, inventing practical codes long before anyone possessed the mathematical vocabulary for what they were doing.

In practice: Code books let customers replace entire stock phrases with single paid code-words, cutting transmission cost by packing more meaning into fewer units.

4

Machines that handle any symbols

Lovelace saw that Babbage's Analytical Engine could manipulate symbols of any kind, not only numbers, separating pattern from physical embodiment more deeply than the telegraph.

Charles Babbage designed the Difference Engine to eliminate human error in mathematical tables and the never-fully-built Analytical Engine, which took input from punched cards borrowed from the Jacquard loom's patterned weaving. Ada Lovelace, translating an Italian article, added notes three times longer than the source and described an algorithm for Bernoulli numbers—the first published computer program. More lasting than the algorithm was her claim that the engine could, in principle, operate on anything expressible in symbols. She suggested it might compose music if musical relations were stated as rules the machine could process. Gleick treats this as the book's conceptual hinge: once a machine follows rules on symbols rather than arithmetic on quantities, any pattern can be encoded, processed, and transformed. Victorian machining precision doomed the hardware; the idea survived as a blueprint for the mathematics Shannon and others later made rigorous.

In practice: Lovelace's notes described a step-by-step method for calculating Bernoulli numbers on the still-unbuilt Analytical Engine and imagined the same machine composing music from symbolic rules.

5

The bit strips away meaning

Shannon's 1948 theory measured information solely as uncertainty reduced, proving reliable codes exist below channel capacity only by ignoring what any message means.

Working at Bell Labs, Shannon published his mathematical theory of communication and defined information by surprise: the less predictable a message, the more information it carries. He borrowed entropy from thermodynamics and adopted the bit—suggested by John Tukey—as the unit that resolves one yes-or-no uncertainty. His own guessing experiments on English showed skilled readers needed few tries for successive letters, implying the language carries roughly one bit of real information per letter even though its alphabet could in principle carry nearly five. That quantified the redundancy drummers and telegraphers already used by instinct. The noisy-channel coding theorem then showed that for any fixed noise level a code exists allowing arbitrarily low error so long as the rate stays under capacity—reliability without infinite padding. Because the math deliberately ignored meaning, the same equations later described telephone lines, computer circuits, and genetic sequences alike.

In practice: Shannon had people guess the next letter in English sentences; high success rates revealed heavy redundancy and roughly one bit of actual information per letter.

6

Feedback holds order together

Wiener's cybernetics reframed stability in machines and organisms as an information problem of comparing output to a goal and correcting, binding Shannon's entropy to physical entropy.

Norbert Wiener coined cybernetics in his 1948 book, generalizing wartime anti-aircraft fire-control research in which machines predicted evasive pilots from incomplete data and continuously adjusted aim. Any system that maintains a desired state—a thermostat, a steering gear, a body regulating temperature—does so through feedback loops that sense difference and act. This recast control as information work rather than pure energy or mechanics. Gleick links it to Maxwell's 1867 demon, the imaginary being that sorts fast and slow molecules and seems to decrease entropy for free. Leo Szilard showed in 1929 that acquiring the needed information itself carries a thermodynamic cost; later Landauer and Bennett refined the point that erasing information generates the compensating entropy. Information was no longer a metaphor borrowed from engineering; it participated in the same physical laws that govern heat and disorder, preparing the ground for finding measurable information inside living cells.

In practice: Anti-aircraft predictors continuously compared a plane's observed path to the gun's aim and corrected, treating incomplete data as an information loop rather than a pure mechanics problem.

7

Heredity as a four-letter message

Once biologists saw genes as transmitted information, DNA's chemistry mattered less than its function as a coded alphabet subject to copying error and repair.

In 1944 Schrödinger speculated that heredity required an aperiodic crystal—a stable molecule whose irregular structure could store large specific information. Watson and Crick's 1953 double helix supplied the molecule; Gleick stresses the deeper functional discovery that its four bases act as a code whose sequences specify proteins. Through the 1960s Nirenberg, Khorana and others cracked the code: each base triplet (codon) maps to an amino acid or stop, the same combinatorial logic as Morse. Biology's vocabulary shifted to transcription and translation, terms taken straight from communication engineering. Mutation became transmission noise; redundancy and error correction became evolutionary necessities. Shannon's tools, built for telephone wires, now applied to genetic stability. Information theory had left engineering and become a general framework claimed by an unrelated science.

In practice: Researchers established that DNA base triplets function like Morse sequences, each codon corresponding to one amino acid or a stop instruction.

8

Culture as competing replicators

Dawkins's 1976 meme proposed that ideas, tunes, and beliefs spread by replication and selection inside minds, though no discrete cultural unit matches a gene's fidelity.

In The Selfish Gene Dawkins argued genes are not the only replicators: catchphrases, melodies, and doctrines compete for mental space and for copying through speech, writing, or imitation. A rumor spreads because it is memorable and easy to copy, not necessarily because it is true—just as a gene spreads by replicating well, not by benefiting its host. The idea itself became a meme, traveling through academic and popular culture until internet usage narrowed it to image macros. Gleick notes the analogy's looseness: a gene has a measurable codon sequence and mutation rate; a joke or doctrine does not decompose into base-pair equivalents and blends in ways genes do not. Evolutionary biologists still debate whether memetics is theory or suggestive metaphor. Gleick presents it mainly as evidence of how far information vocabulary had traveled—from telephone signals to the spread of ideas—inviting both insight and overreach.

In practice: The word meme itself spread for decades and was later repurposed by internet culture to mean image macros, enacting the replication it names.

9

Randomness equals incompressibility

Kolmogorov and Chaitin defined a string's information content as the length of the shortest program that produces it, making maximum randomness identical to maximum information.

Shannon measured information by symbol probabilities. In the 1960s Andrei Kolmogorov and Gregory Chaitin independently defined it by the shortest computer program capable of generating a given string. A million zeros have almost no information because a tiny program prints them; a million fair coin flips usually cannot be shortened at all. A sequence is random precisely when it is incompressible. Gleick rereads Borges's Library of Babel—an infinite library of every possible book of fixed length—as the literary version of the same proof: a collection containing everything carries no more usable information than pure noise, because nothing can be found without already knowing the target. Meaning therefore requires selection and compressibility, not the mere existence of data. That insight pivots the book toward its final claim about the digital flood.

In practice: Borges's library holds every truth and every falsehood; without a way to select, the whole is as uninformative as random noise.

10

Curation replaces capacity

Exponential digital volume did not solve meaning; it made telling signal from noise the central intellectual task, an open problem Shannon's channel math never addressed.

Gleick closes on early-twenty-first-century estimates of annual digital output already measured in exabytes and growing faster than any human can read. Wikipedia stands as a new kind of institution: an encyclopedia built by decentralized volunteers who correct continuously rather than a credentialed staff fixing entries once in print, a relationship between information and authority without clear precedent in the earlier chapters. Every prior technology solved a transmission problem—how to send reliably despite noise, cost, or distance. Solving all of them at once and cheaply produced an unanticipated receiver problem: unlimited volume with no built-in ranking of relevance. Shannon measured a channel's carrying ability and said nothing about a mind's ability to extract sense from excess. Returning to Borges, Gleick insists a world containing every fact does not thereby contain more understanding and may contain less. The book offers no technical fix, only the inheritance: curation, not more capacity, is the work that remains.

In practice: Wikipedia entries are argued over and revised by largely anonymous volunteers in real time instead of being locked by a small expert staff as the OED once was.