Who carries progress

A colleague with a philosophical bent and I had the same argument from opposite ends recently. He distrusts the founders and investors of Silicon Valley, and he holds that the real breakthroughs in science and technology come from academia, and so in the end from the state that funds it. I argued the opposite: progress is carried by people who have to make things work. Entrepreneurs, investors, and companies.

My reasoning goes like this. Someone who has to make a venture survive is selected for the qualities progress needs: being effective, being efficient, daring to try what nobody has asked for. A market does not only reward those qualities once. It trains them, in the people and in the organisations they build, because the ones without them run out of money. An academic can be right for decades without anyone depending on the result. An entrepreneur finds out within a year or two. That feedback loop is the same one I trust when I measure intelligence by what it does rather than by what it knows.

The second half of the argument is the one I care about more. Over the long run, what these people build does not stay with them. Know-how leaks: employees leave and start their own companies, techniques get published or reverse engineered, tools get cheaper until anyone can use them. The sum of many self-interested builders ends up democratising what each of them tried to keep. Large language models are the latest case. A billion dollars bet without a plan became, within a few years, something a single person can run on a laptop.

My colleague has real evidence on his side, and I should say so. The deep learning that OpenAI scaled came out of university labs, Geoffrey Hinton’s in Toronto among them. The web was invented at CERN, a publicly funded laboratory, and Tim Berners-Lee chose to give it away. Where I still disagree is on what counts as the breakthrough. Academia often finds the idea. It usually takes someone with money at risk to turn it into something that reaches everyone, and the Transformer, invented at Google, shows that the line between the two is not as clean as either of us made it sound.

None of this makes me a capitalist in the sense my colleague meant. I am closer to the geolibertarian, or left-libertarian, tradition: Henry George and Silvio Gesell, and in our time Hillel Steiner, Peter Vallentyne, and Philippe Van Parijs. In that tradition you own yourself and what you make, but land and natural resources belong to everyone, and their value should be shared, through a land value tax or a basic income. It fits the argument above. Know-how disperses on its own; land and resources do not, and that is where the sharing has to be designed. And if AI takes us anywhere near post-scarcity, we will have to revisit those designs soon, because the question will no longer be who produces the most, but who owns what is still scarce.

Open threads:

• Is there a fair test of who carries progress, or does each side count the breakthroughs that suit it?
• How fast does know-how actually diffuse, and what slows it down: patents, secrecy, scale, compute?
• What remains scarce when production is cheap, and who should own it?