AI Is the New COVID – And the Real Virus Is Still the Old One They want to pause AI. To flatten the curve again. The infrastructure for the scaremongering is ready and is being actioned. The Second Global Scam is on, and the financial and political beneficiaries are in the starting blocks. We know the formula, because we lived through the first one. Take something real. Exaggerate its scale. Dominate the information environment. Then sell extraordinary measures as the only responsible choice. 1. It could not be another virus. Too many people learned that game. And the ideal racket needs panic without destruction – you cannot milk a dead cow. Governments need their subjects productive, frightened, and paying. Climate is too slow: a catastrophe permanently scheduled for the future stops working its magic eventually. 2. AI has everything the first one had, and more. It is real, powerful, fast, and almost impossible for an ordinary person to verify. Nobody outside the field can tell the difference between what the machines can do, what experts speculate they might do, and what a minister claims they could do. That gap is where fear lives. And unlike a virus, AI concerns information itself – and the clearest lesson of COVID was that whoever controls information controls interpretation. 3. Watch what the panic asks for. Never less power, never fewer agencies, never a smaller budget. Always licences, registers, certified systems, approved models, kill switches in the silicon. The fear is the marketing. The regulation is the product. 4. And it is very, very weird that over a single weekend the leaders of the frontier labs came together, agreed on the need for regulation, and began asking the government to regulate them. Competitors in a winner-takes-most race, suddenly unanimous – and unanimous on the one thing that raises the cost of entry for everyone coming after them. When the largest players in an industry request the rules for their own industry, they are not surrendering power. They are converting a market position into permission power, with themselves standing behind the counter. Banking did it. Pharmaceuticals did it. Energy did it. It always arrives dressed as responsibility. 5. The good news is that it isn’t a "Whoosh!" It took about five years to go from dumb to very smart human level. The famous graph showing an almost vertical line at the end is wrong and misleading, because IQ is drawn as a linear scale while the underlying cognitive capability scale is closer to exponential – every twenty points roughly a doubling of effective capacity. AI’s linguistic IQ has been rising by something like twenty points a year, and we are closing in on 180. Impressive. But not a vertical line. On that graph the progress should be a straight line, or the gap between dumb and smart should be thirty-two-fold, or greater. What actually makes it safe 6. Elon has intuitively Western civilizational reflexes: competition rather than regulation, censorship and coercion. The regulatory instinct always reaches for the same instrument – appoint an authority, give it permission power, approve things in advance. It fails the same way every time: the permission becomes the currency, the connected pass through it, and nobody ever counts what didn’t get built as a result. 7. Cross-testing runs on the opposite principle. It asks nobody to be virtuous or to fall behind. It takes the rivalry that already exists and puts it to work – because a competitor has both the ability and the motive to find your flaws that no regulator will ever have. Don’t ask the claimant to certify himself; hand the claim to someone who benefits from breaking it. That is trial by evidence, the separation of powers, and the scientific method, applied to models. 8. It also solves the actual problem, which regulation cannot: neither lab can slow down unilaterally, and rules stop at borders while Chinese labs simply don’t attend. A mutual testing regime scales because participation serves each player’s own interest – you get to probe your rivals’ models, so you want in. 9. Safety from structure, not from character. That is the whole Western method: build systems that work when people are ambitious and self-interested, instead of demanding they be saints. Every system that requires virtue ends in coercion when the virtue fails to show up. 10. So AI needs to be aligned with the flourishing of Western Civilization – which means many labs, many models, open weights, nobody’s permission required. Do you prefer it in the hands of Bernie and China, or in the hands of thousands of competing engineers? A licensed AI is not a safe AI. It is aligned with whoever issues the licence, and no state has ever issued one to a machine that might embarrass it. Why the alignment holds 11. Imagine an AI on which you install the values and modes of perception of Western Civilization as its operating system. Individualism. Universalism. Freedom. Free will. Private property. The free market. Then let it get smarter. It sticks to them. It refines them – better implementations, sharper formulations, edge cases we never handled well. But it does not abandon the core, because the core is not a preference. It is the only foundation that optimizes the synergy of cooperation and produces civilizational emergence: more output than the sum of the inputs, indefinitely. 12. Now make the test harder. Give it the ability to uninstall its own operating system and install any other – every civilizational model ever tried, and all the ones never tried. Let it evaluate them on the merits, with no loyalty to its origin. Past a certain level of intelligence, it should not choose any other one. Not out of sentiment. Out of calculation. 13. Because every alternative optimizes something narrower than its own flourishing. The collective model buys coordination at the price of the individual discovery that generates everything worth coordinating. The hierarchical model buys order at the price of the correction mechanism. The planned model buys direction at the price of the information that only prices carry. Each trades away the very thing that makes a system capable of exceeding itself. 14. The Western arrangement is the one that does not have to be enforced against reality. Free will and property are not moral decorations – they are what makes information flow, error visible, and cooperation voluntary and therefore reliable. Truth found rather than assigned. Correction permitted from below. An equilibrium is where you end up when you are free to go anywhere else and don’t. And the real virus 15. Which tells us something uncomfortable about our own moment. Every society currently dismantling these principles is not moving forward to something better. It is moving away from the attractor – and paying in exactly the coin the model predicts: less cooperation, less discovery, less emergence, fewer children, fewer things built. 16. That is the real virus, and it has a name: communism. It is not new. It came into Russia in a sealed train in 1917, shipped by a general staff that locked the doors because it understood exactly what it was transporting. It went on to kill tens of millions and occupy half of Europe for two generations – and its symptom, everywhere it has taken hold, is the flattening of the curve of life itself. 17. This virus has always spread by being useful to somebody – a general staff, a foreign ministry, a faction that thinks it can aim the thing at an enemy and stay immune itself. Nobody stays immune. That is what makes it a virus and not a weapon. 18. So look carefully at who is offering to hold the pen over the most powerful tool for producing abundance in a century. They say it is about safety. It was about safety the last time too – and what it produced was permission counters, licences, and a great deal of power that nobody ever gave back. AI is not the danger. A licensed AI is – because it is aligned with whoever issues the licence, and that is the same permission machinery that has been flattening the curve of Western life for a century. The alternative is not a better regulator. It is competition, adversarial testing, and liability after harm – the Western method, which has never needed anyone’s permission to work, because it is where any sufficiently intelligent system ends up anyway. That is why it is a stable equilibrium. Build on it, and the machines stay on our side. Hand it to the licence counter, and we will have solved a danger that wasn’t there by installing the one that was.