Data is not the new oil: a call for a Butlerian Jihad against technocrat data ding dongs

I tire of the dialog on “big data” and “AI.” AI is an actual subject, but as used in marketing and press releases and in the babbling by ideologues and think tank dipshits, the term is a sort of grandiose malapropism meaning “statistics and machine learning.” As far as I can tell “big data” just means the data at one point lived in something other than a spreadsheet.

 “BigDataAI” ideology is a continuation of the program of the technocratic managerial “elite.” To those of you who are unfamiliar with the work of James Burnham, there is a social class of technocratic “experts” have largely taken over the workings of society in the West; a process which took place in the first half of the 20th century. While there have always been bureaucrats in civilized societies, the ones since around the time of Herbert Hoover have aped “scientific” solutions even where no such thing is actually possible. This social class of bureaucrats has had some mild successes; the creation of the American highway system, public health initiatives against trichinosis, US WW-2 production. But they have mostly discredited themselves for decades: aka the shitty roads in America, the unaffordable housing in major urban centers, a hundred million fat diabetics, deindustrialization because muh free market reasons, the covidiocy and most recently, the failure of every noteworthy technocrat in the world’s superpower to predict election outcomes and even its ability to honestly count its votes. Similar social classes interested in central planning also failed spectacularly in the Soviet Union, and led to the cultural revolution in China. There are reasons both obvious and deep as to why these social classes have failed.

The obvious reason is that mandarinates are inherently prone to corruption when there are no consequences for their failures. Bureaucrats are  wielders of power and have the extreme privilege of collecting a pension on the public expense. Various successful cultures had different ways of keeping them honest; the Prussians and pre-Soviet Russian bureaucrats recruited from honor cultures. Classical China and the early Soviets did it  via fear. The Soviet Union actually worked pretty well when the guys from Gosplan could be sent to the Gulag for their failings (or because Stalin didn’t like their neckties -keeps them on their toes). It progressively fell apart as it grew more civilized; by the 1980s, nobody was afraid of the late night knock on the door, and the Soviet  system fell apart when the US faked like it was going to build ridiculous space battleships. The rise of China has largely been the story of bureaucratic reforms by Deng where accountability (and vigorous punishment for malefactors) were the order of the day. Singapore makes bureaucrats meet regularly with their constituents; seems reasonable -don’t know why every society doesn’t make this a requirement. It is beyond question the American equivalent of the Gosplan mandiranate is almost unimaginably corrupt at this point, and the country is falling apart as a result. 

While it gives policy-makers a sense of agency having a data project, consider that there isn’t a single large scale data project beyond the search engine that has improved the lives of human beings. Mind you, the actual civilizational utility of the search engine is highly questionable. What improvement in human living standards has come of the advent of google in the last 20 years? The only valuable content on the internet is stuff made by human beings. Google effectively steals or destroys most of the revenue of content creators who made the stuff worth looking at in the first place. Otherwise, library science worked just fine without blue haired Mountain View dipshits running SVD on a link graph. INSPEC (more or less; dmoz for research) is 120 years old and is still vastly better for research than google scholar. Science made more progress then between 1898 and 2005 or so when google more or less replaced it: and the news wasn’t socially toxic clickfarming idiocy back when the CIA censored the  news instead of google komissars with facial piercings. These days google even sucks at being google; I generally have more luck with runaroo or just going directly to things on internet archive.

If “AIBigData” were so wonderful, you’d see its salutary effects everywhere. Instead, a visit to the center of these ideas, San Francisco is a visit to a real life dystopia.There are thousands of data projects which have made life obviously worse for people. Pretty much all of nutrition and public health research post discovery of vitamins, and statisticians telling people not to drink toilet water is worthless or actively harmful (look at all those fat people waddling around). Most biomedical research is false, and most commonly prescribed drugs are snake oil or worse. Various “pre-crime” models used to justify setting bail or prison sentences are an abomination. The advertising surveillance hellscape we’ve created for ourselves is both aesthetically awful and a gigantic waste of time. The intelligence surveillance hellscape we’ve created mostly keeps its crimes secret, and does nothing obviously helpful. Annoying advertising invading every empty space; I don’t want to watch ads to pump gas or get money from my ATM machine.  Show me something good these dorks have done for us; I’m not seeing it. Most of it is moronic overfitting to noise, evil or both.

It’s less obvious but can’t be stated often enough: often “there is no data in your data.” The technocracy’s mathematical tools boil down to versions of the t-test being applied to poorly sampled and/or heteroskedastic data where they may not be meaningful. The hypothesis under test may not have a meaningful null no matter how much data you collect. When they talk about “AI” I think it’s mostly aspirational; a way out of heteroskedasticity and actual randomness. It’s not; there are no “AI” t-tests in common use by these knuckleheads, and if there were, the upshot wouldn’t look that much different from 1970s era stats results. When they talk about big data, they don’t talk about , or issues like ROC curves and bias variance tradeoff. They certainly never talk about data which is heteroskedastic or simply random, which is most of it. 

In reality, data collection is mostly useless. In intelligence work, in marketing, political work: most of it is completely useless, and collecting it and acting on it is a sort of cargo cult for DBAs, cloud computing saleslizards, technocratic managerial nerds, economists, Nate Silver and other such human refuse. Once in a while it pays off. More often, the technocrat will take credit when things go his way and make complicated excuses when they don’t; just look at Nate Silver’s career for example; a clown with a magic 8-ball.  There’s an entire social class of “muh science” nerds who think it a sort of moral imperative to collect and act on data even if it is obviously useless. The very concept that their KPIs and databases might be filled with the sheerest gorp …. or that you might not be able to achieve marketing uplift no matter what you do… doesn’t compute for some people. 

Technocratic data people are mostly parasitic vermin and their extermination, while it would cut into my P/L, would probably be good for society. At the very least we should make their salaries proportional to (1- Brier) scores; that will require them to put error bars on their predictions, reward the competent and bankrupt the useless. Really though, they should all be sent to Idaho to pick potatoes. Or ….