Claude Shannon / 1916–2001
Mute the world. Build your own.
Claude Shannon turned information into something you can measure, the bit, and then spent the rest of his life building unicycles, juggling machines, and a mouse that learns a maze. This is his life sent through his own 1948 diagram: source, transmitter, noise, receiver, destination.
Constructive dissatisfaction
A slight irritation when things don't look quite right, and the urge to fix them.
Theory × practice
Two degrees chosen out of indecision, one temperament that needed both.
Autotelic
Information theory, unicycles, the stock market: the same game, played for its own sake.
After Fig. 1 of “A Mathematical Theory of Communication”, 1948. This page is that diagram, walked left to right.
Stage 01 / Information source / 1916–1936
Where the message comes from.
Petoskey, Michigan, 30 April 1916. Gaylord, where he grew up. Ann Arbor, where he could not choose between mathematics and engineering, so he took both. Every signal on this page begins here.
The trait
Usefully irritated
Curiosity gets you into the room. It is not what makes the work. What Shannon added, and what I keep finding in the founders and scientists I study, is a second thing on top of it: dissatisfaction. Not the depressive kind. A constructive dissatisfaction, a slight irritation when something doesn't look quite right, and a refusal to leave a clumsy thing clumsy.
He said it plainly in a 1952 talk to colleagues at Bell Labs: the great insights don't spring from curiosity alone. They come from that itch, plus the pleasure of scratching it. His biographers compressed the idea into a phrase I have not been able to shake since I read it, that a genius is simply someone who is usefully irritated.
Shannon is the cleanest example of the combination I have come across. He was curious about everything, satisfied by almost nothing that could be done better, and had no interest in arguing about either.
“I get a big kick out of seeing a clever way of doing some engineering problem.”
Claude Shannon, “Creative Thinking”, Bell Labs, 1952
1916–1936
Two degrees, one indecision
He was born in Petoskey, Michigan, on 30 April 1916 and grew up in Gaylord. At the University of Michigan he could not decide between mathematics and engineering, so he did both and graduated in 1936 with a degree in each. He admitted later that this was not a plan. He simply wasn't sure which he liked best. It was adolescent indecision.
It turned out to be the most useful indecision of the century. Someone content to build things would have stopped at engineering; someone drawn to theory would have stopped at mathematics. Shannon, mechanically and mathematically inclined, needed both, and the two fields were about to merge into one. Communication engineering appealed to him precisely because it blended practice and theory. That blend is his whole personality.
Instrument 01 / Information source
Type the message this page will carry
Every communication system starts with a source that picks one message out of all the messages it could have sent. This one starts with you. The transmitter, the noise, the receiver, and the destination below all work on whatever you write here.
30 characters × 7 bits = 210 bits to send. Printable characters only, up to 40.
The method
Six ways to attack a problem
How does a usefully irritated person actually work? In the same 1952 talk Shannon laid out his methods, and they read like a builder's checklist rather than a mathematician's. He also said the quiet part: it is much easier to make two small jumps than one big one.
- 01
Simplify
Almost every problem is befuddled with extraneous detail. Bring it down to its main issues and you can see what you are trying to do. Excise everything except what makes it interesting.
- 02
Encircle
Surround the problem with existing answers to similar questions, then look at what those answers have in common. Two small jumps are easier than one big one.
- 03
Restate
Change the words, change the viewpoint. Break loose from the mental blocks that come with the way you first saw it, and don't be trapped by the work you have already sunk in.
- 04
Divide
Break an overwhelming problem into pieces. Many proofs arrive by roundabout routes: smaller results that seem to lead nowhere, until you find yourself at the back door of the answer.
- 05
Invert
If the premises won't take you to the conclusion, assume the conclusion is true and see what follows. Try proving the premises instead.
- 06
Generalize
Once you have a result, see how far it stretches. Someone will generalize it eventually, so do it yourself.
The first one is the one I keep meeting again in the founders I study. Elon Musk's version is first principles: take the thing apart until you are left with the physics and the facts you can actually verify, then rebuild from there, deleting every requirement that survived only because somebody once wrote it down. Shannon got there decades earlier and called it simplification. Same instinct, same result: strip a problem of everything except what makes it interesting.
Stage 02 / Transmitter / 1936–1940
Where an idea becomes something a machine can carry.
A transmitter takes a message and turns it into a signal. Between 1936 and 1940, at MIT, Shannon did that to logic itself.
Spring 1936
A postcard on a bulletin board
In the spring of 1936 a job notice typed on a postcard was pinned to an engineering bulletin board at Michigan: MIT wanted a master's student to help run the differential analyzer, at the time the largest analog computer in the world. Shannon pushed hard for it, got it, and called it one of the luckiest things of his life.
The machine deserves a picture in the mind. It was a brain the size of a room, a metal calculus engine of shafts and gears that could grind at a single problem for days and nights on end. One study of the Earth's magnetic field and cosmic rays took thirty weeks of spinning gears. When it finished, it had brute-forced equations no human team could sensibly attack. This was the computer before the digital revolution.
It also came with Vannevar Bush, the man who built it, who would go on to direct American science through the Second World War, and who was the first person to see Shannon for what he was: a near-universal talent whose real skill was not raw calculating power but model making, the reduction of a big problem to its essential core.
1937
Logic you can wire
Bush had locked Shannon in a room with a machine built to automate thought by force, gear against gear. In the middle of that work Shannon realised he knew another way to automate thought, and it was already in the room. The relays and switches that controlled the analyzer could be described by Boolean algebra, the nineteenth-century logic of true and false. An AND was two switches in series. An OR was two switches in parallel. A circuit could compute a proposition.
His 1937 master's thesis worked this out, and it has been called the most important master's thesis of the century, because every logic gate in every digital device descends from it. Less than a decade later the great analog machine was obsolete, replaced by digital computers that did the same work a thousand times faster.
What stays with me is the democratic edge of the idea. Precision logic, like a precision machine, multiplies the power of the gifted and the average alike. You no longer needed a genius in the room. You needed the right circuit. Try it below: the lamp only knows series and parallel.
Instrument 02 / The thesis, on a board
Relay logic: the lamp only knows series and parallel
Shannon’s 1937 insight in one board. Switches in series behave like AND; switches in parallel behave like OR. Close and open them and watch the algebra light the lamp. Every logic gate you own is this, miniaturized.
A · B = 1 · 0 = 0Lamp off
Both switches must be closed for current to reach the lamp.
1938–1940
The genetics experiment, and a flight instructor's letter
Bush distrusted specialization; he thought it was the death of genius, and liked to remind people that being both broad and deep did not end with Leonardo da Vinci or Benjamin Franklin. So he ran an experiment on his student. Claude, do you know anything about genetics? Shannon didn't, not even the vocabulary. Bush sent him off to learn it. The hypothesis, in Bush's own lab-report phrasing, was whether a 23-year-old with no training in a field could produce original findings in under a year. Conclusion: confirmed. Shannon's 1940 doctorate was an algebra for theoretical genetics.
Everyone around him seemed to know what they had. When he took up flying at MIT, his instructor wrote to the university president asking permission to ban him from the cockpit, on the grounds that a near genius of such promise should not risk his life in a crash. The president declined; taking an opportunity away from a young man for being intellectually superior would not be good for his character. Shannon kept flying.
Then the transmitter did its job. Your message, encoded the way a teletype would have carried it, leaves this stage below as a strip of holes.
Leonardo's breadth, on this siteInstrument 03 / Transmitter
Your message, encoded
A transmitter turns a message into a signal a channel can carry. Here each character becomes seven bits, one column of a punched tape: a hole is a 1, no hole is a 0. Nothing about the meaning survives this step, and nothing needs to.
Stage 03 / Channel / noise source / 1941–1945
What the world does to a signal.
The channel is where a signal meets everything that wants to corrupt it. Shannon spent the war years inside that problem, and it left him a bitter taste and a theory.
1941
Freedom from almost the first day
After MIT he chose Bell Labs, then perhaps the foremost technology company on earth. The phone company ran the most complicated communication network in existence, was a complete monopoly, threw off cash, and put a lot of it into research kept deliberately separate from the operating business. The Labs were not asked for clearer phone calls. They were asked to imagine a future in which every form of communication would be aided by machines. In a few decades that building produced synchronized sound in film, early fax and television, wartime radar and sonar improvements, the secure line between Roosevelt and Churchill, the solar cell, the communications satellite and, in 1947, the transistor.
Shannon got what he wanted from it: freedom to work on anything from almost the day he started, and nobody telling him what. He was allergic to administration and to groups. His best thoughts came behind closed doors, in Spartan apartments and empty offices, and most days alternated between the notepad and the clarinet and back.
That clarinet stayed with me, because Einstein, the other great scientist of that era, played the violin. My hunch, and it is only a hunch, is that an instrument clears the fog: while your hands are busy, the brain keeps computing in the background, and the idea arrives afterwards. We have almost no unstimulated time left now, and I think we are losing that power without noticing.
Einstein's version, on this site1941–1945
Fire control is a phone call
The draft terrified him, less for the danger than for the barracks: he could not stand crowds or strangers, and the thought of army life horrified him. He put his mind to work instead, which was also plainly the more useful thing. Almost everyone at Bell Labs moved to war work, and Shannon's assignment was fire control: how to hit a moving target. Picture a gun two storeys tall on a ship pitching in the ocean, trying to bring down a fighter moving at 350 miles an hour. That is the mathematics.
Here the pattern shows again. Nothing he worked on ever went to waste, because he was a man of abstractions who found the analogy between everything. He wrote that there were surprisingly close analogies between the fire-control prediction problem and basic problems in communications engineering. A shell reaching an aircraft and a phone call reaching a listener are both struggles against noise; both need statistical inference; both need machines that turn mathematics into action. Many people would not see the two as related. To him they were the same problem.
The cost
The bitter taste
He hated it. The secrecy, the intensity, the drudgery, and above all the obligatory teamwork. Few of his papers were ever co-authored, and being forced into a team in the middle of a divorce got to him badly enough that he had something like a temporary breakdown. I mention it because the myth of the serene genius is a story the record does not support. He was human, and the war cost him.
Early 1943
Tea with Turing
His other wartime job was cryptography: checking that scrambled speech could be securely reconstructed on the receiving end, work that gave him a window into encoded speech, transmission, and secrecy at once, a combination that may not have existed anywhere else at that moment. It is also how he met Alan Turing, visiting Bell Labs from January to March 1943 on secure-speech work. They got along well enough to meet often, over lunch and tea in the cafeteria.
Asked decades later why he hadn't probed Turing about codebreaking, Shannon's answer was that in wartime you didn't ask too many questions. What they did talk about is what makes the story remarkable from where we sit: machines that would think. Both believed a computer equivalent to the human brain, or better, was possible, and both guessed at ten or fifteen years. They were wrong about the date and right about the direction. This was 1943.
Shannon was obsessed with the idea decades before anyone called it artificial intelligence. His fondest dream, he said later, was a machine that thinks, learns, communicates, and manipulates its environment. He thought machines smarter than their makers were inevitable and called the opposite belief foolish logic. He met von Neumann, whom he called the smartest man he ever knew, and Einstein; he advised the NSA and stayed reluctant to talk about that work for the rest of his life. Now send your message through the noise.
Instrument 04 / Channel and noise source
Send your message through the noise
Think of a bad phone line. Some of what you say arrives wrong: a 1 turns into a 0, a word turns into a different word. Drag the noise slider up and watch your message fall apart.
Then do what you would do on a bad line: repeat yourself. Switch on redundancy and every bit is sent three times, so the receiver can take the best two out of three. It costs three times as much room and it rescues almost everything. That trade — pay in space, get back accuracy — is the whole of error correction, and the readouts below tell you whether it is working.
Decoded at the receiver
US␣FUL^G␣QdIR#NOT`␣Y0MQIL$GO␣L
- Bits put on the wire
- 210
- Damaged on the way
- 18
- Still wrong after repair
- 18
- Letters that came out wrong
- 15 of 30
- What this line can carry C = 1 − H(p)
- 0.60 bits per use
About 8 in every 100 bits are arriving wrong, and a single wrong bit is enough to turn a letter into a different letter. Switch on redundancy to trade space for accuracy.
Stage 04 / Receiver / 1945–1948
Decoding: eight years of napkins.
A receiver reconstructs the message from the signal. Shannon spent the better part of a decade doing exactly that to the idea of information itself.
1948
Before Shannon, after Shannon
Information existed before Shannon the way objects had inertia before Newton. But there was almost no sense of information as an idea, a measurable quantity, something fitted out for hard science. Before Shannon, information was a telegram, a photograph, a paragraph, a song. After Shannon, all of it was the same stuff: a sequence of bits. Any message, from any source, for any recipient, with any meaning, could be represented as bits, and the bit was the fundamental unit.
“A Mathematical Theory of Communication” appeared in the Bell System Technical Journal in 1948, and it is hard to overstate what grew out of it: the theoretical floor under the internet, data compression, error-correcting codes, digital telecommunications, and a good part of the machinery behind today's AI. Everything you have ever sent has been treated the way that paper says information should be treated.
In plain words
What information theory actually says
Strip away the mathematics and the whole thing rests on one move: information is surprise. If I tell you the sun rose this morning, I have told you nothing, because you already knew. If I tell you it snowed in Istanbul in August, I have told you a great deal. The value of a message is how much uncertainty it removes.
So how do you measure that? Shannon's answer is the game of twenty questions. Think of a number between one and eight and I will find it with three yes-or-no questions: is it in the top half, is it in the top half of what's left, and again. Three questions, eight possibilities. That is what a bit is — one yes-or-no answer, the smallest possible piece of news — and it is why the size of anything digital is quoted in bits.
The move that changed the century was noticing that this works for everything. A word, a photograph, a heartbeat, a song, a voice on a wire: they all reduce to a list of yes-or-no answers, and once they do, the machine no longer cares which is which. That is why the same cable carries your voice, your bank balance, and a film, and why a computer can store all three in the same way.
One deliberate omission runs through all of it, and it is easy to miss: Shannon threw meaning out on the first page. Whether a message is a love letter or a stock quote is, he said, irrelevant to the engineering problem. The job is to reproduce at one point what was chosen at another. Everything the internet does sits on that decision to stop caring what the message says.
The consequence
Why your files compress and your calls survive
Two everyday things fall straight out of that idea. The first is compression. Real messages are repetitive: English puts a u after almost every q, an empty sky in a photograph is the same blue ten thousand times over. Repetition is predictable, predictable means unsurprising, and unsurprising means it carries little information — so you can throw most of it away and rebuild it later. That is a zip file, a JPEG, an MP3.
The second is error correction. Every real channel is noisy: a scratched disc, a weak signal, a wire near a motor. So you spend some of that same redundancy on purpose, adding a little predictable padding so the receiver can spot what arrived wrong and repair it. Say a word three times down a bad line and let the listener take the best two out of three — crude, but it is the whole idea in miniature.
Shannon's theorem is the surprising part. Every channel has a top speed, its capacity, and he proved that below that speed you can push the error rate as close to zero as you like, and above it you cannot. Not “engineers will get better at this”: a hard ceiling and a guarantee, in the same result. If you slid the noise up a few screens ago and watched the capacity readout fall, that was this theorem, and the triple-repetition switch was you spending redundancy to buy accuracy back.
The measure
Entropy: how surprised are you?
The heart of the paper is a single quantity. Information is surprise. A fair coin toss carries one bit, because before it lands you genuinely cannot say which way it will go. A coin that comes up heads nine times in ten carries much less, because you already mostly know. Shannon's entropy, H = −Σ pᵢ log pᵢ, is that idea written down: the average surprise of a source, in bits.
Apply it to English and something useful falls out. Letters are not equally likely, and after a q you can bet on a u; the language is redundant. In a 1951 experiment Shannon had people guess the next letter of a text and estimated that English carries only about one bit of information per letter, somewhere between 0.6 and 1.3, far less than the seven the tape above spends on each one. That gap is why compression works and why error correction has room to work: you can throw away redundancy you don't need, or spend it deliberately to survive noise. Measure it yourself.
Instrument 05 / Receiver
Entropy meter: how much is actually being said?
H = −Σ pᵢ log₂ pᵢ. Feed it a source and it reports the average surprise per symbol, in bits. Equally likely outcomes score highest; a source you can predict scores low. Try the coins and the die, then your own message.
25 symbols, 13 distinct. By letter frequency alone this text carries about 3.57 bits per character (the ceiling for 13 symbols is 3.70), against the 7 the tape spends. Shannon's guessing experiment showed real English is closer to one bit, because context makes the next letter predictable. Everything above that line is redundancy: what compression removes, and what error correction spends.
The aftermath
Painful to write up
Two things about what happened next. He knew the work was good and never doubted it; he trusted his own judgment so completely that he would not argue about it. And having solved the problem to his own satisfaction, he had little interest in the part that brings acclaim. Writing things up, he said, was painful. The inner scorecard was the only one he read.
The world's scorecard filled up anyway. He became a scientific celebrity; Bell Labs offered carte blanche, MIT wanted him on any terms; the biography I read this from says he is to communication what Einstein is to physics. He did not use any of it to expand his network. If anything he closed himself off further, ignoring letters and committees, spending his attention on whatever puzzle interested him most. He never argued with people who didn't believe in his ideas. He ignored them and kept working.
Stage 05 / Destination / 1949–2001
Where the message lands.
The destination is the person or thing the message was for. Shannon's own destination, after the paper, was a workshop full of machines, a stock ticker, and a parade he designed himself.
1956
A change of scenery
Having done his pathbreaking work by thirty-two, he could have spent the remaining decades as a public intellectual. He spent them tinkering. Bell Labs did not want to lose him: come in, work from home, do both or neither, stay on the payroll. He left for MIT anyway, and the Labs, knowing how little he cared about money for its own sake, kept paying him for a while, so briefly he was on both payrolls. His reason was restlessness: after fifteen years in one institution he felt himself getting a little stale, and a change of scene and colleagues, he said, is very stimulating.
The workshop
Machines are proofs
What came out of the MIT years was his most whimsical work: an electronic mouse called Theseus that learned to solve a maze, chess-playing machines, a calculator that computed in Roman numerals, the world's first wearable computer, a fleet of customized unicycles, a trumpet that shot fire when played, a machine that solved Rubik's cubes, a chairlift from the porch down to the lake, and years of serious study of juggling. He described all of it as happily pointless and said he had spent lots of time on totally useless things.
He made no distinction between his interest in information and his interest in unicycles; they were moves in the same game. What other people called hobbies he treated as experiments, models that filed a problem down to its barest interesting form. And he was so convinced of a future run by machines, at a time when nothing like today's computers existed, that he was willing to be laughed at to bring it closer. Seen from that future, which is our present, his machines were not hobbies. They were proofs. Three of them are on the shelf below.
The toy shelf / three proofs
Machines are proofs
Working models, not replicas: each one demonstrates the claim its original made. Play with them; the biography continues below.
Theseus 1950
A mechanical mouse driven by relays under the maze floor. On its first run it bumps and backtracks until it finds the goal; on the second, it goes straight there. The relays remembered.
A machine can learn from experience. In 1950.
The Ultimate Machine Early 1950s
A box with a single switch. Flip it on and a hand emerges, flips it back off, and withdraws. Built from an idea of Marvin Minsky's, it does exactly one thing.
A machine whose only purpose is to refuse. Even that can be built, and it still makes people grin.
Juggling 1970s–1980s
He juggled, built juggling machines, and studied the cascade seriously enough to write a paper on it: the trade between time in the air and time in the hand.
Play, taken seriously enough, turns into a theorem.
Toy 01 / Theseus, 1950
A mouse that learns the maze
The real Theseus was a magnet-driven mouse steered by relays under a 25-square board. First run: search. Second run: straight to the goal, because the relays remembered which way worked from each square. Run it.
Theseus is at the entrance. Nothing is stored in the relays yet.
Toy 02 / The Ultimate Machine
A box whose only purpose is to turn itself off
Marvin Minsky had the idea; Shannon built it in his machine shop. Flip the switch and a hand comes out, flips it back, and goes home. That is the whole product. Try to make it do anything else.
Off. It has only one job.
Toy 03 / Juggling, taken seriously
The cascade, and the theorem inside it
He juggled, built juggling machines, and wrote a paper on it. His theorem is a bookkeeping identity: follow one full cycle from a ball’s point of view and from a hand’s, and the two counts have to agree. This cascade is timed by that identity: every throw is in the air for the same time, every catch is held for the same time, and the hands alternate.
(F + D) · H = (V + D) · NF flight time of a ball · D dwell time in a hand · V time a hand is empty · N balls · H hands. Here F = 0.62 s, D = 0.40 s, V = 0.28 s, N = 3, H = 2: both sides come to 2.04 s.
Three balls, two hands: every extra ball has to buy its air time from the hands’ idle time. That is the trade the animation is making.
Another puzzle
The market as another puzzle
He did not care about money and became very rich, which is the tell. Money created markets, and markets were puzzles: problems that could be analysed and played out. It became a family game. Betty and Claude followed the market obsessively; their daughter Peggy remembers being taught to read the Wall Street Journal early because her eyes were better than theirs, reading quotes aloud, and later a small personal computer pulling quotes during the day with printouts drifting around the house.
The world at that altitude is small. His MIT friend Henry Singleton built Teledyne, a conglomerate that Buffett and Munger openly admired, and Shannon sat on its board and invested heavily, simply because he had a good opinion of the man; the biography puts that stake at roughly 27 percent a year compounded over 25 years. William Poundstone reports that when Barron's ranked the recent performance of 77 money managers in 1986, Shannon, unmentioned, had beaten all but three of them, working with his wife and an Apple II. Over roughly thirty years, by Poundstone's account, his portfolio returned about 28 percent a year.
With a young Ed Thorp, the man who wrote the book on counting cards, he built the first wearable computer, a cigarette-pack-sized device for predicting roulette, and tested it in Las Vegas in 1961. Thorp's description of how Shannon thought is the best I have read: he thought in ideas rather than words or formulas, and a new problem was a sculptor's block, chiselled down until the solution emerged like an image. It is simplification again, the first of the six strategies, applied to everything.
Kyoto, 1985
History is taught wrong
He kept a couple of dozen prizes in another room and insisted none of them motivated him. He disliked travelling to collect them; he liked being at home, eating the same things, building gadgets. But he went to Japan for the inaugural Kyoto Prize in 1985, and the lecture he gave there tells you what he admired. History, he argued, is taught wrong, at least in America: most of the time goes to political leaders and wars, the Caesars, the Napoleons and the Hitlers, which he called totally wrong. The important people of history are the thinkers and innovators, the Darwins, the Newtons, the Beethovens, whose work keeps growing in influence.
In the same talk he asked for more engineers. The discoveries of science are wonderful in themselves, he said, but they would not touch the life of the common man without the intermediate work of engineers and inventors, the Edisons and the Bells. Coming from a man who held both degrees because he could not choose, it reads less like a policy position and more like a self-portrait.
The Napoleons, on this siteThe parade
The procession he designed
In a cruel twist, one of the most gifted minds of the century spent its last decade slowly degrading. Shannon developed Alzheimer's, ended up in full-time care, and died on 24 February 2001 in Medford, Massachusetts. But he had already turned his mind to the problem of his own funeral and imagined something entirely unlike grief. He was a joyful, optimistic man; the occasion, he decided, called for humour, and he sketched a grand procession in the Macy's style, meant to amuse, delight, and sum up the life of Claude Shannon.
It runs below in the order he set: a clarinettist at the front, then a jazz combo; six unicycling pallbearers somehow balancing the coffin; the grieving widow; a juggling octet followed by a juggling machine; three black chess pieces carrying hundred-dollar bills, followed by three rich men from the West following the money; a chess float with the British master David Levy playing a live match against a computer; then the scientists and mathematicians, a phalanx of joggers, and a band of 417 instruments bringing up the rear.
The parade / in the order he set
The procession
Mount Auburn
The formula on the stone
He is buried at Mount Auburn Cemetery in Cambridge, Massachusetts. On the back of the headstone, where you have to walk around to find it, is the entropy formula. The man who spent his life on communication and was famously uncommunicative had the equation that created this computational world carved into stone and turned away from the path. I have thought about that a lot. It is, and I mean this precisely, a baller move.
Mount Auburn / the stone
H = −Σ pᵢ log pᵢ
Mount Auburn Cemetery, Cambridge, Massachusetts. The equation that made the information age is on the face turned away from the path, half-hidden by a bush, where you only find it if you go looking. A drawing, not a photograph.
Timeline
One life, in order.
- 1916Born 30 April in Petoskey, Michigan; grows up in Gaylord
- 1936Degrees in electrical engineering and mathematics, University of Michigan
- 1936MIT: assistant on Vannevar Bush's differential analyzer
- 1937Master's thesis: Boolean algebra describes relay and switching circuits
- 1940PhD: an algebra for theoretical genetics
- 1941Joins Bell Labs; wartime fire control and cryptography
- 1943Alan Turing visits Bell Labs; daily tea in the cafeteria
- 1948“A Mathematical Theory of Communication” is published
- 1950Theseus, the maze-solving mouse
- 1956Moves to MIT after fifteen years at the Labs
- 1961First wearable computer, with Ed Thorp, tested in Las Vegas
- 1985Inaugural Kyoto Prize in Basic Sciences
- 2001Dies 24 February in Medford, Massachusetts; buried at Mount Auburn
Destination reached
Your message arrived.
Received at the destination
US␣FUL^G␣QdIR#NOT`␣Y0MQIL$GO␣L
- You sent
- USEFULNESS IS NOT MY MAIN GOAL
- Cost
- 30 characters as 210 bits
- Channel
- 18 bits flipped at p = 0.08; 18 bits still wrong after decoding
Every message you have ever sent, every photo, every call, every token a language model produces, has passed through the same five boxes. Shannon drew them in 1948, muted the world, and built his own.