King Ludd #16: On Beer

Tweet

Not that sort of beer, but Stafford Beer, the brilliantly named eccentric British consultant from the post war era who believed that with the right information network and software, socialist Chile of the early seventies could escape the fate of all other socialist countries. What’s amazing about Beer is just how far ahead of his time he was, and how tempting it is to keep remaking his mistakes, which arguably includes most of the thinking leading to the promotion of Big Data and the Internet of Things.

Seemingly typically for his generation, Beer grew up in the UK, and ended up posted to India in WWII, during which he rose to captain and developed an interest in management, which led him into the then fashionable field of cybernetics. Cybernetics derives from the same Greek word as “government,” which is no coincidence as it’s primarily concerned with the functioning of regulatory feedback loops within systems. This usage was popularised by MIT professor Nobert Wiener whose work Beer consumed and further developed in directions slightly different from his American peers. This in turn led to his involvement in Chile, after which he appears to have temporarily given up material things and headed to rural Wales for some time before eventually landing in Toronto, living in both until his death in 2002 at the age of 75. During all of this he sported a beard of a scale normally left to the Russians, and is said to have lived largely on whisky from a hip flask, chocolate and cigars.

The most interesting phase of his work, thanks to a chance to put into practice many of the ideas which would otherwise probably never happen if enough people were discussing them, was what followed his invitation to Chile after the election of the socialist Popular Unity coalition led by Salvador Allende in 1970. During the initial phase of Allende’s presidency there was a period of mass nationalisation, greatly beyond the regulatory capacity of the government, and this led to the Chileans reaching out to Beer as some had noticed overlap between his ideas and how they thought the Chilean model of socialism would work. Of course it didn’t end that way, and Allende wouldn’t see 1974, thanks to a coup on 11th September 1973 by now internationally notorious General Pinochet. Much can be made of the Nixon administration interfering with the economy through a trade blockade (and worse – Nixon told the CIA to prevent Allende becoming president after the election results indicated that’s what the Chileans had voted for), but similarly the threat of nationalisation successfully pushed out significant private industry, and consequently it wasn’t long before the people were in the streets banging their pans together as what had started as a decent effort at wealth redistribution rapidly went off the rails.

Beer had the task during this supposedly non-revolutionary revolution of building a mechanism which could balance the economy, all under one roof, with decisions made and acted on for the people, by the people. The surprising thing is just how close the team came to pulling it off.

Allende interviewed Beer prior to giving the go-ahead for what became “Project Cybersyn,” during which Beer characterised superficially similar in intent Russian efforts at the same goal as absolut(e) crap. Curiously both men shared this view, but not for the same reasons. For Beer this was based in an argument that a centralised system, as the Russians were using, could not handle so many different variables and thus would represent a meaningless simplification to those pulling the levers, while for Allende having so few people pulling such critical levers of power was politically unacceptable. The idea of Cybersyn was that some nebulous amount of self determination could be pushed to the edges, i.e. factories, and they could largely deal with each other, with government left to handle the large strategy, which in their socialist context meant attempting to increase production, regardless of any market for the produce or the price it would get. To do this a certain amount of information needed to get from the periphery to those in government.

Aside from problems with the mere idea, the Chilean government at the time had only four computers, three of which were almost always busy, and no way to obtain more due to the trading blockade and a rapidly diminishing stockpile of foreign currency, much of which had been used on more urgent matters. This left the project with one computer with which to form a network up and down a narrow but very long country. Luckily for those involved another branch of government had hundreds of Telex machines stockpiled in a warehouse which could be repurposed, and so it was decided that these would be sent out to factories and other nationalised businesses from which statistics would be sent to a central location where the information would be transferred to punched cards, processed, and then somehow turned into a lot of material for display in this room:

This is probably the iconic image from Project Cybersyn, and is of the central control room. The idea was that those managing the economy, by which they meant some vague combination of governing elite and regular factory workers, would hang out in this room, access information which would appear on panels (driven by slide projectors) and make decisions based on the best information available. All those projections were hand produced by designers before being photographed, making it an incredibly labour intensive process.

However, the most interesting aspect of this whole system is the relative sophistication of the modelling the software developers on the project ended up producing. What are now fairly standard inference techniques were applied en masse to a large number of time series in order to attempt to understand the nature of a change. For example, it would try to establish if the recent change in a variable an alteration in the status quo, random aberration, or a sign of linear or exponential growth, and present this along with the raw information. This analysis was also intended to prevent fraud on the part of factory managers — a problem that had plagued their Soviet equivalents.

By current standards the number of variables is trivial, but it’s worth noticing how close the intent is to modern trends. Because of the human steps introducing costs into collecting the data those involved did work to identify the critical variables upfront, though it’s quite clear that if they had had the equipment they would have adopted the record-everything-and-worry-about-it-later approach which is where we are today. The crucial part is the faith that given enough information the optimal conditions can be found, and thanks to the fact any non-trivially sized company internally starts to resemble a Soviet style economy many company leaders seem quite susceptible to this same delusion. By contrast, markets have a habit of surfacing information through prices which does not find itself represented in variables attached to a specific enterprise, but is often highly relevant. A recent example would be stock markets globally dipping with concern about the potential impact of failing to deal with the Ebola outbreak.

This picture, from a 1989 Toyota engineering manual, shows an amazing amount of the applications we’re looking forward to thanks to the Internet-of-Things. (The rest of the book is absolute gold, especially sections on fuzzy logic, which has long since gone out of fashion). Between this and Project Cybersyn it should be obvious that all this is actually a very old idea, and while each time it is explored slightly more of the potential is realised we do not make nearly as much progress on solving the problems as the noise would suggest, even if the complexity of the underlying technology has absolutely exploded. Not long after this somewhat optimistic view of the Japanese led future was produced the bubble burst.

It’s very easy to be cynical about such well meaning ideas, and it is important to keep such views in check if progress is to be made, but at the same time when constructing any analytical system it is necessary to understand what it is that you might be missing which will prove crucial. This is doubly true for cases when the decisions made affect variables which feedback into the next round. In the case of Project Cybersyn that proved to be the political situation, which couldn’t be so easily distilled into a time series suitable for statistical extrapolation.

Nigel Birkenshaw runs Atomirex.