# Aivo Vaske: Bits move fast, atoms much more slowly > Source: > Published: 2026-08-17 09:45:00+00:00 # Aivo Vaske: Bits move fast, atoms much more slowly The spread of artificial intelligence and automation does not necessarily mean that human labor will simply disappear; rather, it shifts to areas where using humans remains cheaper or more flexible than using machines, writes Aivo Vaske. A programmer and designer friend recently wondered how to distribute wealth in a world where artificial intelligence does almost all the work. The question is justified, but major pressure for redistribution is unlikely to arise soon, because AI can rapidly reduce the value of digital work, while automating the physical world still requires enormous capital, energy, materials and time. Claims that artificial intelligence will take over the world and do all the work are somewhat exaggerated, because bits move fast, but atoms much more slowly. In general, people consume food, material goods and a wide range of services from hairdressing to healthcare and entertainment. Even with services, the line between material and immaterial is blurry, because many rely on buildings, equipment, transport, energy and other physical capital. The spread of artificial intelligence and automation therefore does not necessarily mean that human labor disappears; rather, it shifts to places where using humans remains cheaper or more adaptable than machines. Sometimes the service can be very simple or even strange. Varlam Shalamov, for example, describes figures in the Gulag's criminal hierarchy who had other prisoners perform the most trivial personal services. Economically, the principle is similar: when human labor is cheap and machines are expensive, people are used. Automating the physical world differs significantly from automating purely informational work. Software can be copied with almost zero marginal cost, but an excavator, truck or robotic arm cannot. Every physical machine requires steel, copper, semiconductors, energy, factories, transport and maintenance. In other words, transforming the physical economy requires rebuilding physical capital. Therefore, even very rapid development in artificial intelligence may transfer into the real economy much more slowly. The world has an enormous stock of capital such as cars, trucks, tractors, factories, warehouses, ports and construction machinery. The speed at which this can be replaced depends not only on the existence of new technology, but also on its price, reliability, access to capital, infrastructure and regulations. In some wealthy economies with high labor costs, automation may progress quickly, while elsewhere humans will remain cheaper and more practical for a long time. However, the emergence of a fully universal, self‑replicating robot is not a necessary precondition for major economic change. It is enough if increasingly long production chains become autonomous and robots manufacture robot components, automated mines produce raw materials, autonomous transport moves them to factories, and machines build and maintain much of the required infrastructure. The smaller the human role in such a chain, the faster physical capital can begin to grow. New bottlenecks then emerge. If human labor no longer significantly limits production, energy, raw materials, land, production capacity and capital become more important. Robots do not eliminate physical constraints, because even a fully automated economy must mine ore, smelt metal, produce electricity and move billions of tons of materials. It is therefore useful to distinguish between two revolutions. Automating informational work can happen very quickly, because spreading it requires relatively little new physical capital. Automating the physical world may receive a huge boost from AI, but its pace inevitably depends on how quickly we can build more machines, energy production and infrastructure. A world where autonomous machines can, with minimal human involvement, design, supply and construct the next Rail Baltic‑scale infrastructure project is not necessarily science fiction. But the path there is likely much longer than the path to artificial intelligence capable of designing such a project. Bits move fast, atoms much more slowly. -- Editor: Kaupo Meiel, Argo Ideon