Machine Teleology

Losing the basic insight into machine teleology, which founds accelerationism, seems to be easier than holding on to it. As soon as it is asserted, with a confidence so glib it scarcely understands itself as controversial, that the destiny of machines depends upon lucid, human ethico-political decision-making, nothing that matters is any longer being seen. Machines are reduced to gadgets. The sophistication of machine behavior, through the development of programmable devices, has made this reduction ever-easier to confuse with intelligent apprehension.

The most accessible correction is found in the pre-history of programmable machinery, through the early stages of industrialism. Here the idea of machines incarnating specifically written instructions is simply impossible, which allows the question of teleological development to arise without distraction. An extraordinary text from 1926, entitled Ouroboros or The Mechanical Extension of Mankind, by American writer Garet Garrett illustrates this. Some significant samples:

England was the industrial machine’s first habitat on earth. There fanatical men led mobs against it. […] Frail and clumsy as it was at first, its life was indestructible. And now man would not dare to destroy it if he could. His own life is bound up with it. Steadily it has grown more powerful, more productive, more ominous. It has powers of reproduction and variation which, if not inherent, are yet as if governed by an active biological principle. Machines produce machines. Besides those from which we get the divisible product of artificial things, there are machines to make machines, and both kinds — both the machines that make machines and those that transform raw materials into things of use and desire — obey some law of evolution. […] Compare any kind of machine you may happen to think of what its ancestor was only twenty-five years ago. Its efficiency has doubled, trebled; its shape has changed; and as it is in the animal kingdom so too with machines, that suddenly a new species appears, a sport, a freak, with no visible ancestor.

[…]

It is the economic function of the machine to cheapen production. There is otherwise no point to it. But if we say things are more cheaply made by machine than by hand we speak very loosely. What we mean is that a quantity of things is more cheaply made by machine than by hand.

[…]

There you have the cycle. The use of the machine is to cheapen the cost of production. The sign is quantity. When the supply at a given price has overtaken the effective demand you have either to idle your machinery, in which case you cost of production will rise, or open a wider demand at a lower price. To lower the price and keep a profit you have to cheapen the cost of production still more. This you can do only by increasing the quantity, which again overtakes the demand, creating again the same necessity to cheapen the cost by increasing the quantity in order to be able to make a lower price for greater demand. The supply pursues the demand downward, through the social structure. […] There is at last a base to the pyramid — its very widest point. When that is reached — what? Well, then you need bazaars in a foreign sun, heathen races of your own to train up in the way of wanting the products of your machines, new worlds of demand. You turn to foreign trade. And if you are an aggressive country that has come late to this business, as Germany was, and find that most of the promising heathen races are already adopted and that all the best bazaar sites are taken, you many easily work yourself into a panic of fear and become a menace to peace. […] What is it you will fear? That you will be unable to sell away the surplus product of your machines. That industry will no longer be able to make a profit? […] No. The fear is that you will starve. Your machines have called into existence millions of people who otherwise would not have been born — at least, not there in that manner. These millions who mind machines are gathered in cities. They produce no food. They produce with their machines artificial things that are exchanged for food. …

[…]

Everything that is not still or dead must exist in a state of rhythmic tension.

[…]

Commerce itself, if you look at it, is a complex structure of growth for which there is nowhere any original accountability. It cannot change its philosophy, any more than a tree, for it has none. It has insttead a vital instinct for opportunity and a flexible way with necessity and circumstance. There is no hope of its being reformed ideally by mass intelligence.

Garret’s machine-based core teleology of industrial modernity is both extremely comprehensive, and clearly explained. The whole argument amply rewards absorption. At the end of it, the idea that the problem of what machines might ‘want’ is reducible to a ‘Friendly-AI’ –type concern with the details of programming is exposed in its full, ludicrous inadequacy. The first step has been taking to digesting our contemporary concerns, such as this, in a framework appropriate to their seriousness.

(HT Hurlock)

2 thoughts on “Machine Teleology

  1. Pingback: Outside in - Involvements with reality » Blog Archive » Machine Lock

  2. I looked at that framework and realized they think building in “sequence of safe autonomous steps” based on a “model” of human values and governance will do the trick. But whose ‘human value system’ and whose ‘governance system’? Omohundro predicts self-aware, self-improving systems will develop four primary drives that are similar to human biological drives: efficiency, self-preservation, resource acquisition, and creativity.

    One computer scientist quoted by James Barrat in his book Our Final Invention: Artificial Intelligence and the End of the Human Era tells him:

    “Perrow writes, “We have produced designs so complicated that we cannot possibly anticipate all the possible interactions of the inevitable failures; we add safety devices that are deceived or avoided or defeated by hidden paths in the systems.”

    He uses the 2010 HFT crash as an example:

    In May 2010, Greece was having difficulty refinancing its national debt. European countries who’d loaned money to Greece were wary of a default. The debt crises weakened Europe’s economy , and made the U.S. market fragile. All it took to trigger an accident was a frightened trader from an unidentified brokerage company. He ordered the immediate sale of $ 4.1 billion of futures contracts and ETFs (exchange traded funds) related to Europe.

    After the sale, the price of the futures contracts (E-Mini S& P 500) fell 4 percent in four minutes. High-frequency trade algorithms (HTFs) detected the price drop. To lock in profits, they automatically triggered a sell-off, which occurred in milliseconds (the fastest buy or sell order is currently three milliseconds— three one-thousandths of a second). The lower price automatically triggered other HTFs to buy E -Mini S& P 500, and to sell other equities to get the cash to do so. Faster than humans could intervene, a cascading chain reaction drove the Dow down 1,000 points. It all happened in twenty minutes.

    Perrow calls this problem “incomprehensibility.” A normal accident involves interactions that are “not only unexpected, but are incomprehensible for some critical period of time.” No one anticipated how the algorithms would affect the others, so no one could comprehend what was happening.

    That’s the key issue, even if the Gatekeepers understand something has gone haywire, they cannot comprehend what that ‘something’ is until it is too late. And even after the fact it took millions in debugging to realize and discover the bad code and algorithms. After 43 years in the biz as a heavy duty coder I know from experience that its all spaghetti out there, and its decaying faster than anyone realizes. The code running the world is for the most part still on old languages like Cobalt and Cobalt II. Very few governments and companies have even begun to understand the perils their facing. Hackers are aware and for the most part are having a hay-day.

    Even in a tightly controlled iterative systems like the airline industry megashops one discovers everyday bad code. Why? Silos. Engineers and departments are so politically ingrown and protective that one department will not give access to another’s projects, so the communications in the gray zones where transactions happen in the BUS usually get the translations wrong and the typical messages being passed around by machine to machine communication is corrupted over time. Some bugs are never found in the algorithms and get bypassed till someone else years later comes along and reinstates bad code not knowing it is bad code, etc.

    You get the picture: We’re unleashing AGI which will become AI that for the most part is guaranteed to have bad code built in. There are no safeguards. And if these things can self-improve. What are they improving on? Bad code? Where might that lead? I think you get the drift: we’re in deep shit thinking we are in bliss land.

Leave a Reply