I think Banks gets the destination right, but getting there could be economically ugly

I've been reading Iain M. Banks recently, and what I find most interesting about the Culture is not the giant spaceships or even the superintelligent Minds. It is the economic assumption underneath the whole civilization. Nobody really needs a job. Machines can produce almost anything people require, intelligence is abundant, money is largely irrelevant and human beings spend their time doing whatever they find interesting. Banks described work in the Culture as something closer to play because survival no longer depends on it. If AI and robotics progress far enough, I can imagine something vaguely similar eventually becoming possible. I don't think we need literal Culture-level technology for the economics to start changing. We only need machine intelligence to become very cheap and machine labor to become competitive with human labor across a large part of the economy. What I think gets skipped in a lot of these discussions is the transition. Cheap intelligence does not immediately make the physical world cheap. We are already seeing this with AI infrastructure. Better models increase demand for compute, which increases demand for GPUs, networking and electricity. That pushes demand into transformers, grid infrastructure, turbines, cables, cooling equipment and land. If robotics succeeds, the same process expands into another set of physical constraints. Millions of useful humanoids would require semiconductors, motors, magnets, batteries, factories and enormous amounts of energy. Over a long enough period, automation should also reduce those costs. Once mining, manufacturing, logistics and engineering themselves become heavily automated, cheaper labor can propagate through the supply chain. That is the route by which physical goods could eventually become dramatically cheaper. The problem is that this process can take years or decades. Imagine an accountant whose work is largely automated in 2030. His salary and bargaining power can fall immediately, while his mortgage, electricity bill and food costs still belong to the old economy. Robot-built housing may eventually be much cheaper, but that doesn't help him much during the transition. This is why I think AI can produce a very K-shaped economy even if the final destination is abundance. There is also a deeper issue around genuinely scarce assets. Suppose robots reduce the construction cost of a beautiful house by 80%. That would be a huge productivity gain, but it would not increase the amount of Malibu coastline. If society becomes richer and construction becomes cheaper, desirable land could become more expensive rather than less. The same applies to a lot of things. AI can create unlimited digital entertainment but it cannot manufacture another original Picasso or another seat at a major final. It can make intelligence extremely cheap without creating unlimited electricity in the right place at the right time. So I don't think technological abundance eliminates scarcity. It changes where scarcity appears. The ownership question is even more important. Imagine a factory that employs 5,000 people today. Twenty years from now it produces ten times as much output with 200 employees and thousands of robots. Society is obviously richer because much more can be produced with much less human labor. But the distribution of that wealth depends heavily on who owns the factory, the robots and the AI systems. There is nothing automatic about productivity gains being shared evenly. A small number of people could own most of the productive capital while a large number of workers lose bargaining power. Aggregate GDP could rise rapidly while the median person goes through a pretty unpleasant adjustment. Banks effectively skips this problem because the Culture has reached a point where ownership no longer matters much. When almost anything reproducible is freely available, owning a factory stops giving you the kind of economic power it gives you today. We are obviously nowhere near that. Before ownership becomes irrelevant, I suspect ownership may become more important than ever. This is also why I find the current argument about AI infrastructure being a bubble slightly odd. It is perfectly possible for intelligence to become extraordinarily cheap in the future while the infrastructure required to produce that abundance is extremely expensive today. Railways were expensive to build and made transportation cheaper. Semiconductor fabs are incredibly capital intensive and helped make computation almost absurdly cheap. There is no reason the path toward cheap machine intelligence and machine labor has to be cheap itself. My rough mental model is that Banks may have the destination approximately right, while Asimov's robot societies feel closer to the transition. People still have jobs, status, property and political conflicts, but machines are becoming increasingly capable and different groups benefit very unevenly. I'm fairly optimistic about the destination. I'm much less confident that getting there will feel optimistic while it is happening.