(i.e.: the S-curve knows everything except what matters)

…Note to those who skipped previous issues.

This is the third instalment of a column on my bad teachers – those who saw before the others and whom the mainstream only mentions after the damage is done. Ellul had told us that Technology has become an autonomous system that no longer serves man: it uses him. Kaczynski had told us that these systems compete for their own survival without regard for consequences – and he called it self-prop system, not ‘ecosystem of innovation’, which sounds better but says the same thing. Today comes Virilio. Who adds the detail that was missing: the mechanism. That is, the speed. That is, the shipwreck already included in the ship.

The navigator and the lake (story of an invention that worked)

Every year, dozens of motorists around the world drive into lakes, rivers and ditches following the directions of their satnav. They are not stupid. They are not distracted – well, maybe a little, but that is not the point. They follow instructions. They perform correctly what the navigator was designed for. The navigator says: turn right. They turn right. The problem is that on the right, where there used to be a road, there is now water. The cartography is out of date. But trust in the algorithm is. The sat-nav was designed to eliminate the need to think about where to go. It worked. It also invented a new category of accident: one where you stop thinking about where you are going.

Today that navigator no longer just tells you where to turn. It starts deciding where it makes sense to go. And the difference is only subtle at first. Because as long as he takes a wrong turn, we can correct him. When it gets the destination wrong, we can’t.

Paul Virilio had seen it before GPS existed. We will come back to that in a moment. First, let’s talk about the S-curve. Which is correct, accurate, well documented, and describes the wrong problem with almost touching precision.

The curve that is right (and the problem of being right)

There is an S-curve analysis of artificial intelligence out there that deserves to be read carefully, because it is one of the most honest and documented things I have seen on the subject. The author has experienced three booms and three collapses first-hand – dot-com, biodiesel in Brazil, cannabis in Colorado – and knows what he is talking about. This is not the usual advice with perfect diaphragms and optimistic arrows.

The thesis is simple and correct: every industry follows an S-curve. Emergence, growth, compression, plateau. High margins in the initial phase attract capital, capital brings competitors, competitors compress prices, and in the end only the most efficient or the most protected survive. It happened with the Internet. With biodiesel. With cannabis. It will happen with AI, but faster. Much faster. The conclusion: prepare wisely. I agree. Prepare wisely.

The problem is not the curve. The curve exists, it works, it is empirically verifiable. The problem is what the curve does not say – and cannot say, because it is an instrument of description, not prophecy. The curve tells you where you are. It does not tell you what you have already invented while running.

The S-curve is perfect. Like a navigator who takes you into the lake with millimetric precision. It does not make mistakes. It simply does not see the water.

The third bad master (the one who designed bunkers and thought about speed)

Paul Virilio was born in Asnières in 1932. An architect by training, he studied with Claude Parent and designed churches. Then he began to look at the World War II bunkers on the French Atlantic coast – these abandoned concrete structures, half buried by sand – and saw something in them that no one else had seen: the architecture of military speed. The forms that war had produced. The landscape that technology had carved into the land. From there he developed dromology (the science or logic of speed): the study of speed as the structuring force of history. It is not war that shapes the world. It is the speed of war. It is not technology that transforms society. It is the speed of technology. And from there comes the phrase that every engineer, every technologist, every consultant presenting a slide on disruption should tattoo on the palm of their hand:

“The invention of the ship is also the invention of the crash. The invention of the plane is also the invention of the crash. The invention of the nuclear power plant is also the invention of Chernobyl.”

Every technology also invents its own accident. Not as a side effect, not as a bug, not as an unforeseen failure mode. As a structural consequence, internal, already inside the invention at the moment of its birth. Virilio called it the production of the accident. The accident does not happen despite the technology. It happens through technology, and often because of its proper functioning.

The airbag killed the children by working exactly as designed. The navigator takes cars into the lakes by following instructions exactly. Chernobyl exploded during a safety test designed to make the plant safer. The problem was not the malfunction. It was the functioning.

Ellul had told us that technology pursues its own autonomous logic. Kaczynski had told us that self-prop systems compete for survival regardless of consequences. Virilio adds the mechanism: it is speed. Speed is the way in which systems get out of control. Not because they become bad. Because they become faster than those who should stop them.

Not because they become autonomous. But because they become faster than responsibility. And when responsibility lags behind, it does not disappear. It distributes itself until it becomes unrecognisable.

The wall of sound and silence (or: the visionary who hears feedback)

In 1974, the Grateful Dead built the Wall of Sound. Six hundred speakers. Twenty-six thousand watts. A sound system so perfect that it could be heard with crystal clarity three hundred metres from the stage. Each instrument had its own channel, its own stack, its own line. No feedback. No distortion. The cleanest sound ever heard at a rock concert.

It was an absolute technological triumph. It was also so expensive to assemble, dismantle and transport – seventy-five tons of equipment, seventy-five people to handle it – that it almost drove the band into bankruptcy. After a year of touring, the Grateful Dead stopped. They stopped playing live for eighteen months. The Wall of Sound had invented the perfect gig. And it had invented its own silence. But if the Grateful Dead seems too niche an example for you to reason about economics, let me add a more comforting (!!) case for those who believe in numbers.

The two Nobel laureates and the fund that could not fail

In 1994, Long-Term Capital Management was founded. The founders include Myron Scholes and Robert Merton – who won the 1997 Nobel Prize in Economics, in large part for the mathematical models used by LTCM itself. The fund applies quantitative arbitrage strategies of an unprecedented sophistication. The returns in the early years are extraordinary: 40% in 1995, 41% in 1996. The models were correct. The mathematics were flawless. The Nobels were deserved.

In 1998, after the Russian crisis, LTCM loses USD 4.6 billion in less than four months. The Federal Reserve has to orchestrate an emergency bailout because the fund’s failure risks dragging the entire global financial system down with it. Fourteen banks eventually step in. The fund is liquidated. The models had an implicit assumption: that markets behaved according to normal statistical distributions. It worked perfectly, until it didn’t work. And when it stopped working, it stopped so fast and so big that there was no time to correct.

The technology – in this case mathematical and computational – had worked exactly as designed. It had also invented an accident that no single operator could have produced on his own: a global systemic risk built piece by piece by rational, individual decisions, all correct. Two Nobel Prizes, one shipwreck. Virilio would have said: of course. He was already inside the ship.

What the curve does not see (or: the navigator driving himself into the lake)

The S-curve tells you that AI will compress innovation cycles from decades to quarters. It is true. That is exactly what is happening. What the curve does not tell you is what this acceleration invents as it runs.

If the dot-com bubble took four years to inflate and two to deflate, the damage could be seen in real time. There were signals. There were quarterlies. There were journalists starting to ask uncomfortable questions. The incident had a human duration – long enough to be recognised, at least in retrospect.

If the AI compresses this cycle to four months, the accident has a duration that is no longer human. Not because humans cannot understand it. But because the bureaucratic, regulatory, financial and decision-making systems we use to manage accidents are not designed for that speed. The point is not that AI will make mistakes. The point is that it will make them at the right speed so that we don’t see them. We are not building smarter systems. We are building systems that make mistakes faster than we can see them. And when error becomes faster than awareness, it stops being error. It becomes a system.

When Challenger exploded in 1986, it took thirteen months to realise what had gone wrong. Today, with cycles compressed by AI, the next company is already replicating the same mistake while you are still picking up the pieces of the previous one. Speed doesn’t just compress growth. It compresses learning from the incident. And a system that cannot learn from its accidents fast enough to avoid the next one is not a system that improves. It is a system that scales error. And for the first time, the system does not just scale the error. It begins to generate the conditions for the next one itself. It is no longer just speed. It is autogenesis.

Virilio had written this in 1977, in Vitesse et Politique: speed is not a neutral instrument. It is a form of power. He who controls speed controls reality. And when speed exceeds the human capacity to process it, power ceases to belong to anyone and begins to belong to itself. Kaczynski called it self-prop system. Ellul called it autonomy of technology. Virilio calls it dromology. Three names, three angles, one object: something that moves faster than the person who built it. AI does not just speed up processes. It accelerates the distance between the decision-maker and what happens. It is the ultimate phantom traffic jam: everything seems to be flowing, but nobody is driving any more.

Icarus was not stupid (mythological conclusion obligatory)

Daedalus builds wings. He builds them well – mythology does not say that the wings were a faulty prototype. Icarus knows how to use them: his father explained it to him. He flies. He approaches the sun. The wax melts. The standard interpretation: human hubris. The hybris that the gods punish.

Virilian interpretation: the wings work exactly as designed. The problem is that they work so well that they allow Icarus to reach an altitude that wax cannot withstand. The accident is not a deviation from the design. It is the design taken to its logical consequences.

Daedalus had invented flight. He had also invented the fall. Not because he was negligent. Because there is no flight technology that does not already contain, in its structure, the category of the fall accident. The S-curve tells you: watch out, here comes the compression. It’s like saying to Icarus: watch out, it’s hot up there. True. Accurate. Insufficient. What is missing is: the wings already contain the sun. Not the astronomical sun. The sun as the structural limit of technology – the point at which the system meets what it was not designed for, and the wax begins to melt.

The AI is building the most beautiful wings ever seen. The S-curve tells us that the edges will compress. Virilio tells us that the sun is already inside the wings. The question is not whether we will fall. The question is whether we will have enough altitude to realise the fall before we land.

Ellul had told us to look at the system. Kaczynski told us that the system feeds itself. Virilio tells us that the system accelerates. Three bad teachers. Same blind spot of the mainstream. Same answer that nobody wants to hear: it is not enough to understand the curve. You have to understand what has already been invented while you weren’t looking.

– – – – – – – – – –

P.S. – The navigator leading into the lake never stops recalibrating the route. He does not say “sorry, I made a mistake”. He says, “in two hundred metres, dive.”

P.P.S. – Virilio died in 2018. He did not see ChatGPT. He has seen enough.

P.P.S. – The Grateful Dead’s Wall of Sound is on display at the Smithsonian Institution. It is a museum piece. Like almost all great technological inventions, it works best stationary. LTCM, on the other hand, is not on display anywhere. But the mathematical models that blew it up are still in use.