That time I was nearly burned alive by a machine-learning model and didn’t even notice for 33 years

Remember Red Plenty, Francis Spufford’s historical SF novel about the Soviet Union’s efforts to create a real-time planned economy using computers and the ideas of Oskar Lange and Leonid Kantorovich? Sure you do if you’re on this blog. Well, it turns out that it had a dark and twisted 1980s sequel.

We already knew about Operation RYAN, the Yuri Andropov-inspired maximum effort search for intelligence offering strategic warning of a putative Western preventive war against the Soviet Union, and that it intersected dangerously with the war scare of 1983. We also knew that part of it was something to do with an effort to assess the intelligence take using some sort of computer system, but not in any detail. A lot more documents have just been declassified, and it turns out that the computer element was not just a detail, but absolutely central to RYAN.

At the end of the 1970s the USSR was at the zenith of its power, but the KGB leadership especially were anxious about the state of the economy and about the so-called scientific-technological revolution, the equivalent of the Revolution in Military Affairs concept in the US. As a result, they feared that once the US regained a substantial advantage it would attack. The answer was to develop an automated system to predict when this might happen and what the key indicators were.

Model the whole problem as a system of interconnected linear programming problems. They said. Load up the data. They said. Comrades, let’s optimise. They said.

In all, the RYAN model used some 40,000 data points, most of which were collected by greatly increased KGB and Joint GRU field activity. It generated a numerical score between 0 and 100. Higher was better – above 70 peace was probable, whereas below 60 it was time to worry. The problem was the weighting applied to each of those parameters. Clearly, they had to train the model against some existing data set, and the one they chose was Nazi Germany in the run-up to Operation BARBAROSSA.

Who needs theory? They said. We’ve got the data. They said. A simple matter of programming. They said.

As Sean Gallagher at Ars wisely points out, this is a case of the problem described here, that gave us those amazing computer dream pictures. The neural network that classifies cat photos must by definition contain enough information to make a random collection of pixels catlike, although uncannily not quite right. Similarly, RYAN picked up a lot of unrelated data and invariably made it vaguely Hitler-y.

The score went through 60 as early as 1981. The Soviets responded by going on higher alert and sending more agents to posts in the West to get more data. Meanwhile, in the West, John Lehman’s maritime strategy was being put into effect, causing the US Navy and its allies to operate progressively closer to the Soviet periphery, which only made things worse. In the autumn of 1983, the score may have fallen below 40, around the time Stanislas Petrov did his thing.

At this point, Communist Party local cadres were being called in to be briefed on the coming war and their duty to prepare the population. Tactical nuclear weapons were released to local control and moved about by helicopter. The Soviet military was on a higher state of alert than even during the Cuban missile crisis. Fortunately, at this crucial juncture, Yuri Andropov resolved the situation by dying and therefore denying the Big Algo that crucial parameter: patronage.

So, when I was reading all that SF as a kid, I had actually narrowly escaped being vaporised with nuclear space rockets by an evil computer that had convinced itself I was Hitler! I had no idea!

Less flippantly, one of the major themes in Red Plenty is the tension between Kantorovich’s vision of a decentralised, instantly responsive socialist economy, and the Party’s discretionary power – between communism and the Communists, if you like. The RYAN story flips this on its head. This time, it wasn’t the bureaucrats’ insistence on clinging to power that was the problem. It was the solution. The computer said “War”; only fundamentally political, human discretion could say “Peace”. As Joseph Weizenbaum put it, a computer can decide but it cannot choose.

Another thing from Red Plenty that comes up here is that the same unvarying forces of Soviet politics worked the same way, computers or no computers. In the end, everything was personal, and settled through the backstairs gift-economy of favours and influence. Only the loss of its patron could stop the machine.

Also, another theme in the book is the future role the actors in it will play in the perestroika years. We have the cadre down in Novocherkassk who refuses to get used to violence. We have the cadre and programmer who may be turning into an embarrassing trendy dad, but has been enduringly influenced by the Czech experience of 1968. We have the economist who has learned the lesson that the system will have to change dramatically, even if this gets put off 20 years. When they reach the peak of their careers, something is going to change.

And of course they were arriving there just in time to “sudo killall -9 ryand” before ryand killed us all.

17 Comments on "That time I was nearly burned alive by a machine-learning model and didn’t even notice for 33 years"


  1. «The score went through 60 as early as 1981.»

    Amazing coincidence! 🙂

    «Meanwhile, in the West, John Lehman’s maritime strategy was being put into effect, causing the US Navy and its allies to operate progressively closer to the Soviet periphery, which only made things worse.»

    Well, “Strangelove” was a thinly veiled *documentary*: there were quite a few neocon factions that were pressing in the USA policy establishment for a final solution to the communist problem, and never mind for the well documented infiltration of “christian” dominionists in the USA air force and in USA cabinet-level positions. Especially in the 1980s.

    If the USSR just read the neocon policy papers of the 1980s and Reagan’s lips they would have not needed any inference engine to deduct that it could rapidly become USA policy to attack them.

    «In the autumn of 1983, the score may have fallen below 40, around the time Stanislas Petrov did his thing.»

    Just like Arkhipov did in the 1960s during the Turkish missile crisis, when as part of a pattern of provocations and acts of aggressive war, the USA navvy attacked USSR nuclear missile submarines with the transparent purpose to force them to launch:

    «The Soviet military was on a higher state of alert than even during the Cuban missile crisis. »

    This is the usual vile right wing propaganda: in large parts of the world that is known as the Turkish missile crisis. Some very important factions of the USA elites had managed to get dozens of first-strike, short-flight-time, “decapitation” nuclear missile batteries next to the russian border in Turkey, and were trying to get the same nuclear missile batteries installed in various central european countries. People even in right-wing countries in Europe were absolutely terrified by this, because the USA plan was starkly clear, and it was to sacrifice those countries in a first strike against the USSR:

    * To destroy in a surprise attack thanks to short flight time all the strategic nuclear weapons command centers in the USSR, making the USA (relatively) safe from a long range retaliation attack (submarine based missile are very unreliable, and were extremely unreliable then).
    * Then the surviving tactical forces of the USSR would retaliate short-range against the countries they could reach, turning Germany (and the UK and Turkey) into radioactive glassland. Two birds with one stone! 🙂 The New American Century Project version 1.

    Eventually the western european governments insisted on strict safeguards on dual-control, which involved western european troops prepared to fight USA troops if they tried to actually use the nuclear missile batteries for a first attack onto the USSR. This situation eventually persuaded the German elites that the USA elites had powerful factions which were effectively insane and became the motivation for the Ostpolitikl

    This left the turkish missiles which were pretty much under the sole physical control of the USA troops. Indeed during the Turkish missile crisis J Kennedy seems to have been *strongly* advised to turn a crisis into an opportunity by firing those missiles on Moscow and all the relevant strategic command centers.

    The mere installation of those missiles, never mind their arming and readying for launch, was an aggressive act of war by any definition, even worse than theUSA naval blockade of Japan in the 1940s, and the USSR leadership who were not insane warmongers unlike the John Birch society members on the other wise decided on a very proper and diplomatic answer, to re-establish some mutual-assured-destruction by doing something similar but vastly less threatening in Cuba purely as a limited response.

    Now what actually happened is uncertain; some official transcripts show that J Kennedy was utterly surprised by the mention by the USSR of the threat of the turkish first-strike nuclear missile batteries, and this hints that “somebody” in the very top USA military and political elites funded and enabled the commission of an act of aggressive nuclear war without the knowledge of the USA president, which is a case of grave treason, which was surely investigated by B Kennedy. However within a few years entirely coincidentally both J and B Kennedy “died” and nothing came of that investigation.

    The USA officer who was the firing officer for those first-strike nuclear missile batteries has written his memory of those times, as a patriot would, and they are starkly clear:

    http://www.forbes.com/sites/markhughes/2012/10/05/author-recounts-world-on-the-brink-of-war-in-my-turkish-missile-crisis/
    https://events.d.umn.edu/Alworth-Reading—My-Turkish-Missile-Crisis-A-Memoir-from-the-Frontline-of-.htm

    Reply

    1. This is the usual vile right wing propaganda: in large parts of the world that is known as the Turkish missile crisis

      I am fairly sure this is nonsense. Recall that at the time, no link was made public by either side between the withdrawal of R-12s from Cuba and Jupiter IRBMs from Turkey and Italy. In fact, the linkage was only expressed even in private in the back-channel talks, not even in the main diplomatic channel, and the timing of their removal was arranged so as not to give that impression. So how “large parts of the world” would have been able to call it the Turkish missile crisis at the time is beyond me.

      Also, Robert Kennedy wouldn’t have had to do much investigating to find out the Jupiters were there; he was the guy who agreed with the Soviets to remove them, as the US end of the back-channel.

      Reply

    2. The mere installation of those missiles, never mind their arming and readying for launch, was an aggressive act of war by any definition, even worse than the USA naval blockade of Japan in the 1940s,

      This sentence has a very high ratio of crazy to word count. No, there is no definition under which deploying some of your troops in an allied country by mutual consent is an ‘act of war’ against a third country. And the US naval blockade of Japan in the 1940s started after Pearl Harbor. What there was before Pearl Harbor was an embargo, which is very different. For a start, it’s legal.

      Reply

  2. I feel like baconmeteor’s underlying message is that we’re failing to choose a lot these days, letting the “computer” or the “algorithm” of the situation decide instead.

    Reply

    1. But what’s the alternative? Rely on folk sayings? Sheep livers? Sell-side Economists? The writings of long dead prophets?

      Because the risk is not really «the “computer” or the “algorithm”» but the «letting … decide» instead, without questioning it and using is as a support for a call of judgment. Nothing new under the sun.

      In the specific case, the USSR leadership used also spies, analysts and I would dare to say reading the USA newspapers and the transcripts of public speeches of the neocons, all advocating surrounding the USSR with a network of bases capable of launching first strike attacks, funding a massive military expansion in peacetime to provide overwhelming force for an attack against the USSR, and funding and arming many islamic terrorists attacking the USSR as a godless empire. The irony :-).

      And as a barometer the model seemed pretty reasonable to me.

      «The computer said “War”»

      Given the above this statement seems to me quite misleading if not outright evil. The “computer” was not saying “War”, the *model* build by experts embodying expert estimates of probabilities was estimating that USA policy was one of aggression and of preparation for a surprise attack against the USSR, which were indeed very plausible stories, especially given the relatively recent past. Also, while the USSR had formally renounced a first nuclear strike (a policy well communicated to local commanders), the USA has never done so, always reserving the right to “project their might” at any time of their choosing.

      Because American Exceptionalism! As to that a very amusing reminder of what some factions in the USA political system believe:

      http://sedulia.blogs.com/sedulias_translations/2009/05/bush-chirac-gog-and-magog.html
      «In 2003, Thomas Römer, a theology professor at the University of Lausanne [Switzerland], received a telephone call from the Elysée Palace [in Paris, home of the President of France]. Jacques Chirac’s advisers wanted to know more about Gog and Magog…. two mysterious names that had been spoken by George W. Bush as he was trying to persuade France to go to war at his side in Iraq.»
      «Bush is said to have declared to Chirac that Gog and Magog were at work in the Middle East, and the Biblical Prophecies were being accomplished. It was a few weeks before the intervention in Iraq. The French president, to whom the names of Gog and Magog meant nothing, was stupefied.»

      This is reported here directly by T Römer:
      http://www2.unil.ch/unicom/allez_savoir/as39/pages/pdf/4_Gog_Magog.pdf

      And was confirmed by J Chirac in a book-interview with JC Maurice:
      https://books.google.co.uk/books?id=I3Qz8iCh6t8C&pg=PT40

      The USA military elites and large part of the political ones in the 1960s and 1980s were not quite as stable and sensible as GW Bush…

      Reply

      1. Also, while the USSR had formally renounced a first nuclear strike (a policy well communicated to local commanders), the USA has never done so

        I’m sure Vasily Arkhipov would be delighted to hear that it was all a fuss about nothing, because the USSR pinky-promised it would never launch first. Of course, the fact he was in a submerged submarine when he did his thing is the tell here; if you send something to sea you inherently put it in the ultimate authority of the ship’s captain whether you like it or not, and this goes double for a submarine.

        Reply

  3. «Indeed during the Turkish missile crisis J Kennedy seems to have been *strongly* advised to turn a crisis into an opportunity by firing those missiles on Moscow and all the relevant strategic command centers.»

    The article linked above in Forbes, a rather “conservative” publication, casually annotates that:

    «the U.S. temporarily went to Defcon 2 during the crisis, partially due to a power struggle between the president and his generals over precise control of the nation’s nuclear arsenal»

    BTW the other link is dead, but the book is still available:

    http://overcoatbooks.com/books/my-turkish-missle-crisis-a-memoir-from-the-front-line-of-1962/

    Reply


  4. This bit just jumped out at me: “Development of the RYAN computer model began in the mid-1970s, and by the end of the decade the KGB convinced the Politburo that the software was essential to make an accurate assessment of the relationship between the USSR and the United States…The fundamental assumption of RYAN’s forecasting was that the US would act much like the Nazis did—if the “correlation of forces” was decisively in favor of the US, then it would be highly likely that the US would launch a surprise attack just as Germany did with Operation Barbarossa.”

    So the fundamental basis of the USSR’s defence strategy was an assumption that, in any given set of circumstances, Jimmy Carter would act in exactly the same way as Adolf Hitler.
    I mean, I know the Great Man theory of history is frowned on, but this might just be taking things a bit far the other way.

    Reply

      1. If the scoring included parameters for security in Afghanistan, Central America and Southern Africa then the score would start moving towards the HERE BE DRAGONS point within a few weeks of Ronald Reagan being sworn in.

        Reply

      2. “I do like Blissex’s point that the score went through the HERE BE DRAGONS marker as soon as they reached a year that ended in 1:-)”

        Not sure that was the point he was trying to make, but it does have rather a plausible badly-trained-neural-network sound to it…

        Reply



Leave a Reply to ajay Cancel reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.