Elon Musk’s version

Last week, Elon Musk gave an interview to The Economist, conducted by editor-in-chief Zanny Minton Beddoes, which should be listened to whilst undertaking a rarely practised exercise: separating his professional and political roles. Musk’s political views are often reprehensible, his forays into European affairs crass, and his use of X irresponsible. But Musk the politician and Musk the entrepreneur are not one and the same. When he talks about immigration, political parties and culture wars, he can become a confused propagandist. When he talks about rockets, factories, energy, robotics and artificial intelligence, it’s a different story. Not because he’s always right, but because he can envisage seemingly impossible technologies and build companies that force the world to take them seriously.

The crux of the interview – a summary of which you are now reading, written using Il Foglio AI and a lengthy prompt from this newspaper’s editor – lies in the opening question: what will the world be like in 2036 if his projects are successful? Musk does not start with cars, satellites or Mars. He starts with intelligence. He predicts that within five years, AI could surpass the sum of human intelligence and that within ten years, there will be very little that a machine cannot do better than a person. Human beings, he adds, will no longer be in charge. It would be difficult to imagine chimpanzees leading a civilisation more intelligent than themselves; the gap between us and a superintelligence could be even wider. Everything else follows from this: work, wealth, money, energy, China, robotics and space.

For Musk, the revolution lies in the union of digital intelligence and physical intelligence. Today, AI writes, calculates, programmes, analyses and generates images, but remains confined to a screen. To truly transform the economy, it must acquire hands, legs, sensors and tools; it must move from bits to atoms. We need humanoid robots designed as physical terminals of a vast distributed intelligence. The combination of AI and millions of robots can generate an almost unlimited productive capacity: artificial brains that design and make decisions, artificial bodies that build, transport, assist, assemble and repair. AI is not an industry alongside the old ones, but a capability destined to permeate every industry.

Musk describes the outcome as an era of extraordinary abundance, in which production will exceed human needs. The prediction sounds like science fiction, but it urges us not to use the past to measure an unprecedented transformation. The year 2036 should not be taken literally: Musk often gets his timelines wrong and goes on to achieve things later on that were previously dismissed as impossible. The point is the direction. This gives rise to the most provocative prophecy: money may no longer matter. If AI and robots were to produce more goods than we can consume, money would lose part of its function. Beddoes emphasises the transition; Musk tends to leap across the bridge and describe the other side. The insight remains powerful: if the marginal cost of production were to plummet, the problem facing the economy would no longer be how to produce enough, but how to distribute abundance.

The same upheaval applies to work. Musk does not merely predict that certain professions will be replaced. His argument is radical: “I think it will be a bumpy ride, because AI will be capable of doing any job better than any human.” It will happen first in digital activities, “in anything that involves a person sitting in front of a computer or a phone”. In software, he argues, AI is already better than most programmers and could soon reach the ‘Stockfish level’ – the gap in chess between the best human players and an unbeatable programme. The profession would not disappear immediately, but its focus would shift: less code writing, more problem selection, goal setting and solution verification.

For physical labour, the next step will be needed: robots capable of moving about in the real world. When models control millions of artificial bodies, human labour will become optional. The metaphor used is that of gardening. No one needs to grow tomatoes to eat: they can buy them at the supermarket. Yet some people still grow them for pleasure and to share with friends. Work could become an activity undertaken not for survival, but for pleasure, identity, creativity or the desire to feel useful. For centuries, humanity’s challenge has been to free itself from toil; in Musk’s world, it could become a matter of finding meaning in the time thus freed up.

Beddoes asks the inevitable question: how will people make a living? Musk responds with a concept more ambitious than universal basic income: “universal high income”. The Treasury, he argues, could issue cheques. When the interviewer reminds him that creating money without new revenue generates inflation, Musk turns the argument on its head: “If the production of goods and services increases dramatically”, the money supply can grow without inflation, provided it increases more slowly than supply. “The problem will be deflation, not inflation.” Here, the limitation of his vision becomes apparent. There is no teleportation between present-day capitalism and automated abundance. There is a phase in which certain professions disappear, wealth and power become concentrated in the hands of companies that own models, chips and robots, and democracies must decide who pays and who controls. Musk has a formidable theory of supply, but one that is far more fragile than that of institutions.

Musk makes no secret of the extreme risk. In the past, he had estimated the probability of AI leading to a catastrophic outcome at between ten and twenty per cent. He does not retract that fear, but says he sees no realistic way to halt AI and robotics. Even if there were a button capable of stopping everything, perhaps it should not be pressed, because the most likely outcome remains one of extraordinary abundance. He describes the future as “thrilling and terrifying”. Beddoes sums it up as follows: AI is inevitable; we must hope for the best and enjoy the journey. Musk agrees.

However, it does not propose proceeding blindly. Leading AI companies should meet every week or two to discuss security. Before the launch of a much more advanced model, competitors should be given one or two weeks to test it. This would amount to a peer review between rival laboratories: competitors have both the expertise and the incentive to uncover flaws. If several companies deemed a system dangerous and the manufacturer refused to rectify it, the government should intervene. The proposal is debatable, as it entrusts the initial scrutiny to competitors, but it raises a serious point: a bureaucracy lacking technical expertise cannot, on its own, assess systems that change more rapidly than the law.

The other key issue is China. Musk rejects the idea that the United States will automatically retain its leadership. The fundamental constraints on AI are chips and electricity; for physical intelligence, the ability to manufacture robots is also a factor. China is already extremely strong in modelling, robotics and energy. At present, it is held back mainly by access to the most advanced semiconductors, but Musk believes Beijing is close to solving the problem of lithography – that is, the machinery required to produce sophisticated chips on a large scale. If that bottleneck were to be removed, the combination of industrial scale, engineers and energy could rapidly shift the global centre of gravity for AI.

For this reason, he considers it of little use to prohibit American companies from using Chinese models. The United States can regulate its own businesses, but it cannot prevent the rest of the world from adopting competitive technology. You do not beat China by banning a model, but by producing better energy, data centres and chips. This is a lesson for Europe, which discusses AI as if it were primarily a matter of regulations. AI is also about steel, copper, transformers, power stations, grids, permits, capital and construction sites. The limiting factor is not just the availability of processors, but the energy required to power them.

This is where xAI, Tesla and SpaceX cease to appear as separate companies. Whilst on Earth, power and cooling constraints limit computing capacity, Musk envisages data centres in orbit powered by the sun. Starship becomes not merely a rocket, but the infrastructure for transporting vast amounts of computing power into space. Even Mars is not an isolated project. Musk says he wants to “spread consciousness into the future” and extend its reach throughout the universe. AI, satellites, robots and rockets all serve the same purpose: to increase the chances that intelligence will survive, spread and make the universe less empty.

This is the Musk worth listening to, even when he exaggerates or mistakes a possible trajectory for a definite date. Not the oracle, but the entrepreneur who sees connections where others see separate industries: AI needs electricity; computing infrastructure; robotic body models; data centres might need rockets. Tesla, xAI, SpaceX, Starlink and Optimus are not, in his mind, a portfolio of businesses. They are parts of a single machine: producing more intelligence, giving it physical form, powering it and taking it everywhere.

This is what separating one’s professional and political careers means. One can regard Musk as politically unreliable without pretending that he is irrelevant as an entrepreneur. One can condemn his identity-driven obsessions without refusing to listen to him on China, energy and robotics. The opposite risk to cult-like adoration is automatic ‘anti-Muskism’: the idea that, because many of his political views are repugnant, every one of his technological predictions must be nonsense. The crucial part of the interview is not the prophecy about money. It is the method: starting from physical constraints, imagining what would happen if they were removed, and formulating the right question. Not whether AI will change the world, but what will remain unchanged when digital intelligence, robotics, energy and space begin to function as a single system.

“We need to prepare ourselves: AI will be capable of doing any job better than any human being”

In the future of AI, says Musk, a ‘universal high income’ will be inevitable