On Saturday 30 December 1972, in Santiago, the President of Chile sat down in a white fibreglass chair with an orange cushion and pressed a button on its arm. The chair was one of seven, arranged in a ring in a six-sided room panelled in wood, with no table and no paper anywhere. On the wall in front of him were four screens in fibreglass housings, one large above three small; behind the wall, slide projectors were supposed to answer the button by throwing a chart of the nation’s industrial production onto the glass. It was high summer, the electronics in the space behind the panels had been baking for days, and they did not answer properly. “The projectors went crazy,” one of the designers, Rodrigo Walker, said afterwards: you pressed for one thing and got another. Salvador Allende, a physician by training, sat in the chair with his minister of economy, Fernando Flores, and his minister of the interior, General Carlos Prats, and said, by Flores’s account, that it would be good to have it working, and to keep at it.

Eight days later the Observer in London told its readers that Chile was being run by computer. The system in the room, the paper said, had been “secretly brought into operation”. It had not been brought into operation at all. What existed, at the end of December 1972, was a network of about a hundred telex machines in state-owned factories, a statistical program on a government mainframe that read the numbers they sent each day and looked for changes, a half-built model of the national economy, and the room. The whole of the daily input to the machine was, by the estimate of the man who built it, fewer than 5,000 numbers.

Those numbers were the instrument. The experiment was whether a government could read them and, having read them, leave the factories alone.

Who ran it

Chile elected Salvador Allende president on 4 September 1970, with a little over a third of the vote and a margin of 1.3 per cent over the candidate of the right. He took office in November at the head of Popular Unity, a coalition of Socialists, Communists and smaller parties, with a programme of taking the mines, the banks and the largest industrial firms into public ownership by lawful means, and without a majority in Congress. The instrument for the factories was the Corporación de Fomento de la Producción, CORFO, the state development corporation founded in 1939, which now took on the running of every firm the government bought, expropriated or “intervened”. By the end of 1971 the big copper mines and sixty-eight private companies had passed into the state’s hands, by one count, or about 150 enterprises by another; by October 1972 sixty-one more had been requisitioned or placed under an interventor, a manager appointed by the government. Many of the old managers had left, and some had taken their companies’ records with them.

The third-ranking post at CORFO, general technical manager, went to Fernando Flores, an industrial engineer of twenty-eight who had been dean of engineering at the Catholic University in Santiago at twenty-seven and who belonged to MAPU, a small party of young intellectuals that had broken from the Christian Democrats in 1969 and joined Allende’s coalition. The job put him in day-to-day charge of the nationalised factories. From March 1971 he was also president of INTEC, CORFO’s technology institute. As a student in the 1960s he had done piecework for the Chilean team of a French operational research consultancy called SIGMA, which was working for the Chilean steel industry and the railways, and whose head was an Englishman named Stafford Beer.

Beer was born in 1926. He had left a philosophy degree unfinished for the army in the Second World War, served in India, gone into the steel industry afterwards, and in 1950 been handed Norbert Wiener’s Cybernetics, which set the course of his life. He ran operational research for United Steel, then SIGMA, and by 1971 was a consultant teaching at the Manchester Business School, with a large house in the Surrey commuter belt, a beard, and a manuscript called Brain of the Firm, which Allen Lane was to publish in 1972. He had never been to South America; in 1962, when the Chilean steel industry hired SIGMA, he had sent a team and stayed at home.

On 13 July 1971 Flores wrote to him. He was familiar with Beer’s work, he said, and was now, as Eden Medina quotes the letter from Beer’s papers, “in a position from which it is possible to implement on a national scale—at which cybernetic thinking becomes a necessity—scientific views on management and organization.” Beer answered on 29 July that he would surrender any of his retainer contracts for the chance to work on it, “because I believe your country is really going to do it.” A month later the two men met at the Athenaeum in London; Flores had little English and Beer no Spanish, and they managed in a mixture of French, English and Latin. The arrangements were settled in London in September.

He arrived in Santiago at the beginning of November 1971, in the week of the first anniversary of Allende’s government, and was at work within the hour. Flores had already gathered a dozen or so people, most of them his friends, who had been reading cybernetics: Raúl Espejo, a CORFO engineer who would end up running the project; Herman Schwember; Roberto Cañete, a former naval officer who became Beer’s interpreter; Sonia Mordojovich, who coordinated it. At ECOM, the state computer corporation, Isaquino Benadof took charge of the software; a chemical engineer, Mario Grandi, took the economic model; and at INTEC the state’s new industrial design group, under Gui Bonsiepe, a German who had taught at the Ulm school of design, would build the room. In Britain, Beer subcontracted the statistical programming to Arthur Andersen and Company in London and the economic simulator to Ron Anderton at Queen Mary College. Each member of the Chilean team was given the manuscript of Brain of the Firm to read in translation. Beer was paid 500 US dollars a day for his ten days in Santiago, and put up at the Hotel Carrera, across the street from the presidential palace.

On 12 November 1971, after ten days of planning, Beer went alone with Cañete to La Moneda to explain the scheme to the President. The rest of the team waited in a hotel bar. He drew his model on a sheet of paper between them: five nested levels, taken from the human nervous system, with the top level the brain’s cortex. When he came to the fifth, the level he had privately meant to label with the President himself, Allende, as Beer told the story for the rest of his life, smiled and said: “At last: the people.” Allende’s conditions, as Beer recorded them, were that the thing be decentralised, anti-bureaucratic and run with the participation of the workers. The plans were approved. Over the next twenty-two months Beer flew to Chile and back ten times by his own count, eleven by Medina’s.

What they were trying to find out

Beer stated the problem as two questions. In a lecture in New Delhi in December 1974, a year after it was over, he put them this way: “What exactly is going on in all those plants? Secondly how quickly will we know about today’s results?” In an industrial economy, he said, a government’s statistics were “six months or more out of date”, so that governments were “forever reaching precisely the wrong decisions” about putting money into the economy or taking it out. The cure he proposed was not more statistics but faster ones. “We saw no reason why the economy should not move into a mode of regulation operating in real time.”

The hypothesis had two parts, and they were tested very differently. The first was about measurement: that the state of a nationalised industry could be known, not from accounts six months old, but from a small number of figures sent in from every plant each day, if a statistical filter could tell the figures that mattered from the ones that did not. The second was about government: that a centre which knew this could afford to leave each plant free to run itself, and intervene only when the filter cried out, because autonomy, in Beer’s definition, was a system “subtracting at all times as little horizontal variety as is necessary to maintain the cohesion of the total system.” He called the balance point “effective freedom”, and he believed it could be computed. A journalist who sat by the goldfish pond in his study reported him saying that “liberty must be a computable function.”

The machinery of the second claim was what he called the viable system model, the five-level diagram he had drawn for Allende, applied by recursion: a plant inside an enterprise, enterprises inside a sector such as textiles, sectors inside one of four branches, and the branches inside industry, each level the same shape as the one below. At every level a few “critical variables” would be agreed with the people who ran the operation, and for each variable two values fixed: capability, what it should do when everything ran smoothly, and potentiality, what it could do with a little investment. The number that arrived each day was actuality. The ratio of actuality to capability was productivity; of capability to potentiality, latency; the product of the two, performance. Every number in the country would thus be reduced to a figure between nought and one.

The filter was to be Bayesian. For each day’s index the computer would weigh four probabilities: that the movement was chance, that it was a transient, that it was a change of slope, that it was a step. The first two meant nothing happened. Either of the others meant the enterprise was told at once, and only the enterprise. Each variable also carried a weight and a clock, agreed with the plant: how long, given its technology, it would take to put right a real decline. If the clock ran out before the index recovered, “an algedonic cry of pain”, Beer’s phrase, would pass up one level, and then, if nothing was done there, another, so that in theory the President’s economic committee might one day hear of a broken limestone crusher in the north. “Reports back to the enterprises from this monitoring facility were not available to anyone else,” Beer said. “In this lay the devolution of power.”

Notice what the second claim required of the first. The freedom was to come from the fact that the only people who saw a plant’s warning were the people in the plant, and that they acted on it. The experiment’s instrument was therefore not the room, and not the model of the economy, but the daily indices and the exception reports the filter sent back; and the test was whether the managers of some hundreds of newly nationalised factories, many of them political appointees with workers’ committees at their backs, would read them.

What they built

The project was named Cybersyn in English, from cybernetics and synergy, and Synco in Spanish, for Sistema de Información y Control. It had four parts.

The first was the network, Cybernet. Chile had fewer than fifty computers in the whole country in 1968, and ECOM, which owned the government’s four mainframes, three IBM 360s and a Burroughs 3500, could spare time on one of them. Beer proposed that each factory report by telex, the teleprinter network that businesses used for orders and newspapers for copy, and that the messages be punched onto cards and fed to the one machine. “Now there is a wasted facility,” he said of the telex in 1974. “We used it in Chile to provide the economy with a nervous system.” The difficulty was that CORFO owned one or two telex machines and had no foreign currency to buy more. An army major on Flores’s staff, Gustavo Soto, found that ENTEL, the state telecommunications company, had 400 telex machines in a store, bought by the previous government in the 1960s and never installed. Espejo called the store “our Aladdin’s Cave.” The machines were lent to the factories. By March 1972 data were coming into Santiago. By the time of the strike in October 1972 ninety-nine machines had been installed and connected, from Arica in the north to Punta Arenas, over the existing microwave trunk routes, and CORFO had approved buying enough to serve the 500 new subscribers it expected in 1973. The round number that has attached itself to the project since, 500 telexes, matches neither the 400 in the store nor the ninety-nine installed; it is CORFO’s forecast for 1973.

The second was the statistical filter, Cyberstride. Arthur Andersen’s London office agreed to write a “temporary suite” by mid-March 1972 and design a permanent one for Chilean programmers to code; the first price, £34,000, Flores refused, and the work was redrawn at £19,000. The filter rested on a paper of 1971 on Bayesian short-term forecasting by two British statisticians, Harrison and Stevens. Alan Dunsmuir of Andersen carried the temporary suite to Santiago in early March, and in the middle of the month ECOM sent the first results to CORFO. Beer telexed Anderton that the suite “really works too”, and that “the whole thing was impossibles and we did it.” In the factories, meanwhile, operational research teams from CORFO and INTEC were drawing what Beer called quantified flow charts, pictures of each plant with boxes sized to the process and lines as wide as the flow, and picking about five critical variables apiece: stocks in, stocks out, a bottleneck, and in every case absenteeism, which Beer took as a measure of social unease. Each variable’s capability and potentiality were agreed with the management, and a hundred days of data were collected to set its statistical character. By early July 1972 Benadof’s team had the temporary suite checking thirty production indicators. The IBM 360/50 was so busy that the day’s figures could wait twenty-four to forty-eight hours in the queue, so Benadof rewrote the code for the less-used Burroughs 3500. Beer’s claim to a British audience in February 1973 was that a quarter of the country’s industry was reporting daily; the whole input, the New Scientist worked out, was “probably less than 5000 numbers per day,” about ten from each firm, and a manager who had done nothing wrong would typically get one report a week.

The third was the model of the economy, CHECO, for Chilean Economy, which the team sometimes called Futuro. Anderton and Gilligan undertook to write a rough approximation by March 1972 for £2,500, of which £1,000 was computer time at Queen Mary College; it was written in DYNAMO, the simulation language Jay Forrester’s group at MIT had built and which the Limits to Growth model had made famous. Grandi’s team in Santiago tested the model against the national statistics, and the results were poor. Beer blamed the age of the statistics. In 1974 he said plainly that the model’s wall in the room “was never made to work.”

The fourth was the room, the Opsroom. Beer wanted “an environment of decisions” with no paper in it, as unlike a cabinet room as he could make it; the designers at INTEC wanted a circle, could not build one, and settled on a hexagon with five walls for displays and one for the door. Seven swivel chairs stood in a ring, each with a panel of large buttons in the armrest, built for men who could not type, and the buttons worked Datafeed: slide projectors behind the wall throwing charts and photographs of factories onto the back of the screens, so that a man could choose his own path through the pictures. Another wall carried the five-level model, lit from behind; another a display of red lights for the algedonic alarms, which in the prototype had to be posted by hand; another was reserved for CHECO’s projections. The control electronics came late from Electrosonic in England. The prototype was fit to be sat in at the end of December 1972 and, by Medina’s reckoning, working by 10 January 1973. Beer called it a set for a futuristic film, and said so to reporters. The companion piece “The room” sets his description of it against the designers’.

The money, for a national government, was small. By September 1973 Chile had paid Beer nearly £79,000 for the project’s expenses, of which about £33,000 was his own fees; his rate for 1973 was £13,000, paid late. Flores said afterwards that it was a very small sum against what CORFO was spending and what was at stake. Fewer than a hundred scientists worked on it, and the two computers were the state’s, borrowed.

Two thousand telexes

By 10 October 1972, 12,000 lorry owners were on strike against the government, and within days the number was 40,000. The strike had begun over a plan for a state haulage company and had been joined by the retailers’ associations, which shut shops, by the private bus owners in Santiago, seven in ten of whom stopped running, and by the professional guilds of doctors, lawyers and engineers. Lorries did not move food, fuel or raw materials, and the roads were blocked. Beer, who had landed in Santiago on the day the stoppage became national, stayed a week and then had to go back to London.

Flores did not reach for the room or the model. He set up an emergency command in La Moneda, with specialised centres for transport, industry, energy, banking, agriculture, health and supply, a private telephone network of eighty-four numbers linking the people he needed, and the telex machines that were already in the factories for Cybersyn. The Cybersyn telex room at CORFO became the industrial command centre. Alongside the production figures, the machines now carried plain messages. “There were enterprises that reported shortages of fuel,” Espejo recalled to Medina; the room could “distribute this message to the enterprises that could help.” An interventor who was running out of fuel asked his sector committee; one who knew the raw material he needed was in Valparaíso asked for a lorry to fetch it. “It was a very practical thing,” Espejo said. Gustavo Silva, who ran the energy centre, tracked the tankers leaving the depots and whether they arrived: “We knew exactly how many trucks we needed, so each time we lost one we were able to requisition another.” The government had about 200 lorries it could count on against the 40,000 that had stopped, and the network was how it placed them. Beer estimated that 2,000 messages a day passed over the network during the strike; twenty machines in one room, he said, and “the noise was indescribable.” Flores and other senior officials slept in the command centre.

The strike ended on 2 November, when Allende brought the armed forces into his cabinet, with General Prats as minister of the interior, and the military declared it over. Beer wrote later that one senior minister “said flatly that the government would have collapsed that night”, 17 October, “if it had not had the cybernetic tool”, and in 2001 he still described a strike “we absolutely defeated by using computers and telex machines, of all things.” Espejo was more careful: the network “played an important role”, but so did the mobilisation of the neighbourhoods and the factories and the decision to bring in the generals, and “it would be presumptuous to say that the strike ended because of what we did.” Historians of the period credit the last two and seldom mention the telexes at all. Flores, who was made minister of economy as the strike ended, said afterwards that most people in the government never thought of the telex as part of Cybersyn, but as “a very astute idea to produce coordination and communication.” Within a week a colleague, Schwember, was writing to Beer that the growth of their real power and influence had exceeded their greatest fantasy. Within three months Beer was writing to Schwember that his relationship with Flores, which had gone marvellously while Flores was at CORFO, had all but collapsed, because Flores was now a minister with a counter-revolution to think about. “I knew that we could win a lot during the October Strike [by] using my room,” Flores said thirty years later, “but . . . the room was not going to stop tanks and planes and bigger strikes.” Cybernetics, he had come to think, was valuable where your enemy was not trying to kill you.

That was the state of things when the President sat in the chair on 30 December. The team had planned to announce the project simultaneously in Santiago and in Brighton, where Beer was to give the Richard Goodman Memorial Lecture at the Polytechnic on 14 February 1973. Beer had described the system to someone in London on condition of silence until then, and on Sunday 7 January 1973 the Observer ran Nigel Hawkes’s story under the headline “Chile run by computer”. The system, it said, had been “secretly brought into operation” and “assembled in some secrecy so as to avoid opposition charges of ‘Big Brother’ tactics.” Beer landed in Santiago two days later and wrote home that the article had “led to a cabinet crisis while I was there—but don’t tell anyone.” On 10 January Flores announced a plan for the state to control the distribution of thirty staple goods, which the opposition took as rationing; on 17 January the St. Petersburg Times in Florida reprinted the Observer piece with a cartoon of a mainframe watching a small man in a sombrero through binoculars; and on 23 January the centrist Santiago weekly Ercilla, which had learned of the project from the British press, told Chileans about it under the headline “El ‘hermano mayor’ de Mr. Beer”, Mr Beer’s Big Brother. Beer telexed Espejo on 29 January that “silence is a mistake” and proposed a televised tour of the room; the Chileans refused, on the ground that showing it would invite sabotage.

The lecture went ahead on 14 February under the title “Fanfare for Effective Freedom: Cybernetic Praxis in Government”; 300 people came. The next morning the New Scientist carried two pieces by its technology policy editor, Joseph Hanlon. The report, “Chile leaps into cybernetic future”, began: “The dream of operations research — to monitor an entire country’s economy by a real time computer system — is actually happening in Chile.” It gave the two computers by name, the fewer than 5,000 numbers a day, the one report a week, and a sentence of Beer’s on the strike: “communication is indeed control”. The comment on the leader page, “The technological power broker”, described “a computer system monitoring all of the economy, designed by a small group of scientists and imposed from the top in secret”, in which “managers have autonomy so long as they do what the computer expects”, predicted that “managers will soon act to make the computer respond correctly, independent of the actual factory”, and ended by calling Beer “the super-technocrat of them all.” On 15 March, eleven days after congressional elections in which Allende’s coalition gained ground, the Santiago magazine Qué Pasa ran “Plan secreto ‘Cyberstride’: la UP nos controla por computación”, warning that the system would serve “not only for economic ends, but also for political strategy”. The same day the New Scientist printed a letter from Herb Grosch, the American mainframe veteran, who found it “absolutely not possible” that such a thing had been built in Chile; “I call the whole concept beastly,” he wrote. The magazine of the British Society for Social Responsibility in Science read Beer’s algedonic signals as a way of disciplining workers. Within ten weeks the project had been read as Big Brother by a British Sunday paper, a Florida daily, a Chilean centrist weekly, a Chilean right-wing weekly, a liberal science magazine and the socialist left of British science, and as impossible by an American computer man. The companion piece “Fifteen February” reads the British coverage in full.

What it looked like from inside

From Benadof’s office at ECOM it looked like this. The factories’ figures came in by telex each afternoon. His staff punched them, ran the program, and sent the results by telex to the control room at CORFO. They did not, he told Medina, notify the individual interventors, and they did not run the algedonic procedure Beer had designed. The project’s own telexes show that when the filter flagged an index, ECOM alerted the enterprise, the CORFO telex room and Espejo’s informatics directorate all at the same time, so that the autonomy of a plant depended on nobody at CORFO walking down the corridor. Getting the factories to send their numbers every day, Benadof said, was a constant frustration.

From the factories it looked like a visit from young engineers. The modellers were in their twenties; one of them, Tomás Kohn, thought his presence probably annoyed several managers, and recalled a textile manager who was happy to talk about the technical side and not about anything else. Workers were supposed to help draw the flow charts of their own plants, and in most places were not asked; the engineers, Kohn said, dealt with the management and avoided the politics. The training films and booklets that were to explain the thing to the shop floor were planned and not made. Where an interventor took the system up, it was used: at Mademsa, a maker of electrical goods whose interventor was Raimundo Beca of ECOM, the workers took the indices as an argument for bonuses and for counting the plant’s output collectively, and when Beca left the factory the work stopped. The New Scientist said as much in February 1973, in Beer’s own account of what remained to be done: “because the system was imposed from the top in relative secrecy, there is little understanding of it at lower levels. So the government must get managements to use the exception reports and to add their own indices to the system.”

From CORFO, after the strike, it looked like a report. Espejo now ran a new informatics directorate, created to put the project on a proper footing and to end the informal accounting Flores had used to pay for it. Each day the telex traffic from the whole country was collapsed into a single document, with charts and graphs, written at CORFO and carried by hand to the sector committees and to La Moneda. The figures in it were about three days old, where the previous national statistics had been six months old, and they let the government follow the rise and fall of production until September 1973. The report was the one thing in the project that the government read every day, and it was not an exception report from a Bayesian filter. It was a digest of telexes.

From Beer’s side it looked like decybernation. He returned on 16 April 1973 to find the pieces being taken up without the whole. On 19 April Kohn wrote him a long letter: “The cybernetic adventure is apparently coming to an end, or is it not?” Management, he said, “accepts your tools, but just them. . . . The final objective, ‘the revolution in management’ is not accepted, not even understood.” Beer answered on 27 April with a paper he called “On Decybernation”, in which Cybersyn, stripped to its technologies, was “no longer a viable system but a collection of parts”; the result would be “an old system of government with some new tools”, and “these tools are not the tools we invented.” He thought the dismantling “utterly disastrous”. On 21 May, in his annual message to Congress, Allende reported that 26.7 per cent of the nationalised enterprises, accounting for half the revenue of the state sector, had been connected to the system in some degree. Beer’s figure in 1974 was that the picture existed for about three-quarters of nationalised industry by the time the government fell; Medina’s count of the archive cannot match his figures.

From the room, the view was of a prototype. Beer said in 1974 that it “was built and was working. But it was simply a prototype, and the economy of Chile was never actually commanded from there.” Bonsiepe’s description was simpler: “This is a room for making decisions, not for reading reports and talking about notes.” Nobody made a decision in it. Espejo, half a century later, allowed that the room had too much display and no way of sending anything out, and that the criticism was fair as far as it went.

What happened

Beer came back on 20 June 1973 for six weeks, most of them at Las Cruces on the coast, writing a cybernetic model of capitalism which he left with Espejo. On 26 July he saw Allende for the last time; while he waited he talked at length with the President’s naval aide, Captain Arturo Araya, who was shot dead that night. On 2 August, from England, Beer wrote to Allende that the workers in the industrial belts were “inventing their own approach as they go along”, without the models prepared for them, and that this was “a grievous loss of opportunity for the Chilean process”. The letter was not answered.

In August the lorry owners struck again. Espejo, from the informatics directorate, kept the national account by telex: a tenth of the lorries moving at the start, three in ten by the end; food supplies held at between half and seven-tenths of normal; raw materials delivered at normal levels to 95 per cent of the enterprises judged strategic; fuel at 90 per cent of normal with 65 per cent of the tankers. Allende made Flores secretary-general of the government, in charge of its communications; he was thirty. In the first week of September Qué Pasa put the “secret plan Cyberstride” on its cover and reported that the telex network was helping the government seize the factories and had given it “a terrible weapon of control”. On 8 September Allende asked the team to move the operations room into La Moneda; Beer said the order came a week before the President’s death, and that it meant ripping out historic rooms, and that the plans were nearly ready.

On the morning of Tuesday 11 September 1973 the navy rose in Valparaíso and the army and air force in Santiago. Espejo, who had said that everyone knew the coup was coming and nobody knew when, left the CORFO building with Gustavo Toro, the project’s coordinator since June, carrying four packets of copies of the project’s papers, which Espejo kept. By two in the afternoon La Moneda had been bombed by the air force and Allende was dead inside it. A week after the coup an official of the British government called on Beer to establish the facts of Chile’s application, filed early in 1972, for technical aid to reimburse what it was spending in England on his contracts. “Allende was dead,” Beer said; “the work was over.”

The army tried to understand the room. Officers summoned the team and other Chilean computer people and questioned them; Benadof thought they could not grasp a system built to leave the lower levels alone, which was the opposite of what an army is for. One soldier took a knife and put it through every slide the designers had made for Datafeed. Then interest lapsed. Medina’s account is that the military “either abandoned or destroyed” what the team had built and, when it failed to understand the room, dismantled it. Bonsiepe left the country with the only colour photograph of it. By 1975 the junta had adopted the programme of the Chicago-trained economists, which was to sell the firms CORFO had taken, and a system for running a nationalised economy without unemployment had nothing to run.

Flores was arrested on the day of the coup and held for nine months with Allende’s other ministers on Dawson Island, in the far south, then at Ritoque near Valparaíso, where he met Schwember again; he was a prisoner for three years, and went afterwards to California. Espejo went to England and taught at Aston. Beer never returned to Chile. He gave up the house in Surrey for a cottage in Wales, and Maturana said of him that he had arrived in the country a businessman and left a hippie.

After

The system’s parts had afterlives, which is what Beer had feared. Espejo and his students rewrote Cyberstride for microcomputers under the name Cyberfilter and sold it, through a company Beer chaired, to a German chemical firm, a Norwegian aluminium company and an English local authority, and found that what the clients wanted was help understanding their own organisations, not the software. In the late 1980s a team in Uruguay, with Beer’s advice, built a Cyberfilter for the office of President Sanguinetti; Espejo judged it, decades later, a good executive information system and not what Chile had been trying to do, and noted that Beer had lost interest as it became one. Performance indicators that detect change as it happens, Espejo wrote in 2022, became commonplace later.

Beer gave his account to India in December 1974 and in the last five chapters of the second edition of Brain of the Firm in 1981, and his papers, including the telexes, went to Liverpool John Moores University, where Eden Medina read them for a dissertation at MIT, an article in 2006 and a book in 2011. In 2023, for the fiftieth anniversary of the coup, she and two Chilean colleagues built the room again, full size, seventy-two square metres of hexagon with seven chairs whose buttons worked, at the cultural centre under La Moneda, with the original designers helping. The Observer’s headline lives on in the popular telling, where Chile under Allende is still the country that was run by computer.

What it taught

The hypothesis had two parts, and the record on each is different.

The first part, measurement, was built and never read. A daily feed of a few thousand numbers from some hundreds of factories was collected, filtered and reported, and the reports went where the design said they must not: to everybody at once, by telex, from an office that did not use the alarm procedure at all. The managers for whom the exception reports were made did not, for the most part, know they existed, and the workers who were to add their own indices were not told. Where one interventor read the numbers, his factory used them; when he left, it stopped. This is not evidence that the filter was wrong. It is evidence that an instrument is only an instrument to the people who are obliged to read it, and that nobody in the design was obliged. Beer’s own list of what remained to be done, in the week the project became public, began with getting the managements to use the reports.

The second part, government, could not be tested without the first, and so was not. Whether a centre that read the warnings would in fact have left the plants alone is unknowable, and Hanlon’s prediction in February 1973, that managers would learn to make the computer happy independent of the factory, was never given time to come true or false. What was tested, by accident, was something Beer had not written down as a hypothesis. In October 1972 the network built to carry indices carried messages instead: this plant has no fuel, that road is open, send the lorry here. Two thousand messages a day, read by people who had to act on them within the hour, kept a government supplied for three weeks. “Communication is indeed control,” Beer said, and meant it as vindication; it was also the verdict. The part of the system that worked was the part that any government with a hundred teleprinters and a telephone list of eighty-four numbers could have had, and the part that had been sold, in Brighton and in the Observer, as a nervous system for the economy, was the part that did not. Even the one daily document the government read was not the filter’s output but a digest of the telexes, three days old.

The third instrument, the one the press read, was the room. Seven chairs, buttons in the armrests, screens that were slide projectors, a model of the economy on one wall that did not run: it was, as Beer said, a set for a futuristic film, and film is what the newspapers saw. The Observer and Ercilla and Qué Pasa read it as surveillance; the British left as a machine for disciplining workers; Grosch as a fraud; the army as something to put a knife through. All of them were reading the look of the thing and not the fewer than 5,000 numbers a day, which were the thing. The one reading that found the actual weakness, Hanlon’s, found it by asking what a manager would do with a report, which is the question a designer should have asked.

What the record supports is narrower than either the headline or the lament. A government with almost no computers and a store of unused teleprinters built, in four months, a communications network the length of a country, and the network proved its worth in the one way nobody had designed it for. The same government built a statistical instrument on a sound idea and never got its readers; built a room to show the instrument’s output and never commanded anything from it; and set out to model a national economy with a borrowed compiler and a model that failed its first test against the national figures. Between a model of an economy and an economy lay inflation at levels Chile had not seen before, a black market, two national strikes, a hostile Congress, an “invisible blockade” of credit from Washington, and a workforce that was not a factor of production but a crowd of people with opinions; none of these was a critical variable in any flow chart. The experiment was stopped at twenty-two months out of a planned five years by an air force, so that nothing about its third year can be known. The lesson is not that a national economy cannot be run in near-real time on its own measurements. Nobody found out. It is that the measurement was the easy part, that the reading was the experiment, and that the people who built it and the people who feared it both mistook the room for the instrument.

Sources

• Joseph Hanlon, “Chile leaps into cybernetic future”, Technology Review, New Scientist 57, no. 833, 15 February 1973, pp. 363 to 364, and “The technological power broker”, Comment, same issue, p. 347. The two computers, the fewer than 5,000 numbers, the ten critical numbers, the hundred days, the one report a week, the dates of November 1971, March 1972 and November 1972, the “five steps” that remained, the goldfish pond, “liberty must be a computable function”, “communication is indeed control”, “imposed from the top in secret”, the reprogramming remark, and “super-technocrat”. Read from the Google Books scan of the issue mirrored on archive.org. archive.org
• Stafford Beer, Cybernetics of National Development, Evolved from Work in Chile, the Zaheer Science Foundation lecture, New Delhi, 5 December 1974. “At last: the people”, the two questions, the six months, “regulation operating in real time”, the telex as “a wasted facility”, the two computers in Santiago, the triple index, the Harrison and Stevens filter, the algedonic clock, the operations room’s screens, chairs, projectors and slides, “simply a prototype”, the CHECO wall “never made to work”, the 75 per cent, the ten flights, the technical-aid official a week after the coup, and the order to move the room into La Moneda. In copyright; quoted in short phrases. Text as mirrored on archive.org. archive.org
• Eden Medina, “Designing Freedom, Regulating a Nation: Socialist Cybernetics in Allende’s Chile”, Journal of Latin American Studies 38, no. 3, 2006, pp. 571 to 606, read in the Spanish translation by Soledad Ayala and Mariano Fressoli, “Diseñar la libertad, regular una nación”, Redes 20, no. 38, 2014, pp. 123 to 166, which Cambridge University Press licensed. The meeting of 12 November 1971, Flores’s letter and Beer’s reply, the Athenaeum, the 500 dollars a day, ECOM’s four machines, the 48-hour delays, the four sub-projects, the eleven trips, the Andersen and Anderton contracts, the first print-out of 21 March 1972, Benadof’s account of the afternoon routine, the 2,000 telexes a day, the 200 lorries against 40,000, the daily report to La Moneda, the 26.7 per cent, the 8 September request, Kohn’s letter, “On Decybernation”, the Mademsa anecdote, and the room’s dismantling. ridaa.unq.edu.ar
• Eden Medina, Cybernetic Revolutionaries: Technology and Politics in Allende’s Chile, Cambridge, Massachusetts: MIT Press, 2011. The 400 telexes in ENTEL’s store and “Aladdin’s Cave”, the ninety-nine machines and 500 subscribers, the Andersen prices, the strike’s course and the command centres, Espejo’s and Silva’s recollections, Beer’s “would have collapsed that night”, the 30 December visit and the projectors, the Observer’s sentences, the sequence of January 1973, the Brighton telexes, the Qué Pasa and Grosch quotations, the ECOM alerts, Kohn’s and Beer’s papers of April 1973, the letter to Allende of 2 August, the August strike figures, the knife through the slides, Flores’s imprisonment, and the sums paid to Beer. In copyright; quoted in short phrases. The Observer of 7 January 1973 and the Chilean weeklies Ercilla (23 January 1973) and Qué Pasa (15 March and 6 September 1973) are quoted here at second hand from this book, which cites them. mitpress.mit.edu
• Raúl Espejo, “Cybersyn, big data, variety engineering and governance”, AI & Society 37, 2022, pp. 1163 to 1177, open access. The CORFO background, the early team, the manuscript of Brain of the Firm, the four components, “almost real time” and the few days’ delay, the shortcomings of the room, Cyberfilter, Syncho, the Uruguayan project, and Beer’s model of capitalism written at Las Cruces in June 1973. europepmc.org
• Eden Medina, “Designing Freedom: The Cybersyn Operations Room”, Cabinet 46, 2012, an excerpt from the book, for the hexagon, the kitchen wall, the Datafeed screens, the hold button, the 30 December visit, Hawkes’s byline and the St. Petersburg Times cartoon, and the fate of the slides. cabinetmagazine.org
• Stafford Beer, “Fanfare for Effective Freedom: Cybernetic Praxis in Government”, the Richard Goodman Memorial Lecture, Brighton Polytechnic, 14 February 1973, reprinted in Platform for Change, Chichester: Wiley, 1975. In copyright and not read here; its sentences are quoted only as the New Scientist reported them the next morning and as Medina quotes them.
• “Project Cybersyn”, 99% Invisible, episode 230, 2016, transcript, for Flores’s and Bonsiepe’s recorded words, Beer’s 1974 recording on commuting between London and Santiago, and the single colour photograph. 99percentinvisible.org
• John Bartlett, “Allende’s Strange Plan to Connect Chile, Long Before the Internet”, Americas Quarterly, 11 July 2019, for the election margin, Espejo’s “we all knew the coup was coming”, Flores’s and Beer’s later lives, and Maturana’s remark; it misdates Beer’s arrival to 1972 and gives the strikers as 50,000. americasquarterly.org
• Michael Brindley, “Designing a revolution”, MIT News, 16 October 2023, for the 2023 reconstruction of the room, its seventy-two square metres, the original designers and the exhibition at the Centro Cultural La Moneda. news.mit.edu
• P. J. Harrison and C. F. Stevens, “A Bayesian Approach to Short-term Forecasting”, Operational Research Quarterly 22, no. 4, 1971, pp. 341 to 362, the paper on which Cyberstride’s filter was built. doi.org
• Metaphorum, “Cybersyn”, the Stafford Beer society’s summary page, for the four sub-projects and the eudemony meter. metaphorum.org
• The Stafford Beer Collection, Liverpool John Moores University, reference GB 2993 SB, where the project’s telexes, reports and correspondence cited above are held; Archives Hub record. archiveshub.jisc.ac.uk
• Liverpool John Moores University, “Professor Stafford Beer 1926–2002”, for his dates. ljmu.ac.uk