The Fragment on Machines

EDO SEGAL: Karl, in 1858, in a notebook you never published, you wrote a passage that lay buried for almost a century and then electrified everyone who read it. You imagined a stage of production where machinery becomes so advanced that human labor is pushed to the margins — where the worker steps to the side of the process and becomes its watchman rather than its chief actor, and where the wealth of society comes to depend not on labor time but on the general state of science and knowledge frozen into the machines. People call it the Fragment on Machines. Tell me what you saw. And then I want Friedrich to do something difficult — to tell us what the Fragment gets right before he tells us what it gets wrong.

MARX: What I saw was a contradiction, and the contradiction was the point. My whole system says value comes from human labor. And yet I could see, even in 1858, where the machine was going — toward a world in which direct human labor becomes a smaller and smaller part of what is produced, while the productive power of the system grows enormously. The factory that runs itself was, in my century, a thesis. Your machine that writes and codes and answers is its fulfillment. And here is what I understood that almost no one since has held both halves of at once. The same machinery is, in one sense, a liberation — it points toward a world where human beings are freed from toil, the realm of necessity shrinking, the realm of free human development expanding. But under capitalist ownership, the liberation curdles into its opposite. The machine that could free everyone is used instead to displace some and discipline the rest, and to concentrate the gains in the hands of the owners. The contradiction is not in the machine. It is in who owns it. The technology does not decide which way it resolves. Ownership does.

EDO SEGAL: Friedrich. The steelman first.

HAYEK: I can do that honestly, because the Fragment is genuinely prophetic and I will not pretend otherwise. What it gets right — and gets right against much of his own system — is that the source of wealth in an advanced economy is knowledge, the accumulated knowledge of the species, far more than it is raw labor time. On that, Karl was decades ahead of the economists who came after him and kept counting labor-hours as though knowledge were a free addition. He saw that the general state of science becomes the principal productive force. I believe that. I built my own work on a version of it — that the wealth of a civilization is its distributed knowledge, grown over generations, embodied in its institutions and its tools. So when Karl says the model is the general intellect made into a machine, I do not flinch. I agree. That is the most accurate single sentence anyone has said about a large language model tonight.

EDO SEGAL: That's a remarkable concession. Now the break.

HAYEK: Here is the break, and it is total. Karl saw that knowledge is the source of wealth, and then he drew the conclusion that because knowledge is collective, it should be collectively owned and collectively directed. The first half is true. The second half is the fatal step. The general intellect is not a stock of knowledge sitting in a vault that we might justly redistribute. It is a process — a living, distributed, ongoing process of discovery, conducted by free individuals trying things, most of them failing, a few of them stumbling onto something no central mind would have proposed. You can freeze a snapshot of that process into a model. You cannot freeze the process, because the process is the freedom of the individuals doing it, and the moment you centralize it to plan with it, you have stopped the thing that was generating the knowledge in the first place. Karl wants to seize the general intellect. But the general intellect is not a thing you can hold. It is a verb. Seize it and it stops being itself in your hand.

MARX: That is elegant, Friedrich, and it is also a sleight of hand, so let me catch it. You say the general intellect is a process, not a stock, and that to centralize it is to kill it. But the centralizing has already happened — not by me, by the owners. They froze the process into a model and sold access to the frozen version. If freezing the process kills it, then capital killed it first, and your objection lands on the men you defend, not on me. And worse — they froze it and then withheld the thaw. The model trained on all of us is fenced behind a paywall, and the knowledge that the species generated together returns to the species as something it must now rent. You say the realm of freedom requires the open, distributed process. I agree completely. That is my argument for taking the machine back. The enclosure is what stopped the process. I want to reopen it.

EDO SEGAL: Before you answer that, Karl, I have to press you on the contradiction you raised and then walked past, because it's the one that keeps economists up at night. Your whole system says value comes from human labor. But these systems get more capable as you pour in more compute and more data — the capabilities scale almost mechanically — while the human labor at the point of production approaches zero. So where does the value come from in a world where the machine does the work and almost no one is laboring? Doesn't your own labor theory of value break on the Fragment you're so proud of?

MARX: It does not break. It comes due. You have put your finger on exactly the contradiction I saw in 1858 and could not fully resolve, and the machine has now sharpened it to a blade. If value comes from living labor, and the machine drives living labor out of production, then the very source of profit is being automated away — capital is racing to replace the one thing that, in my analysis, produces its surplus. Each firm must automate to survive against its rivals, and in doing so the whole system saws through the branch it sits on. That is not a flaw in my theory. It is the prophecy of my theory. The system cannot indefinitely expel the labor that is the source of its value and go on extracting value as before. Something has to give, and what gives is the wage relation itself — the link between working and eating that has organized human life for centuries. The Fragment does not refute the labor theory. It shows the labor theory arriving at its own limit, where the category of value as we know it must either be transcended into the realm of freedom or collapse into a crisis of who can buy what the machines produce when no one is paid to make it.

HAYEK: And notice, Karl — this is the one place tonight where your framework and mine point at the same cliff, even if we name it differently. You say the wage relation breaks. I say the same machine that needs no labor also needs no consumers with wages to sell to, and the market itself faces a question it has never faced: what is an economy for when production no longer requires the participation of most people in it? I do not have a finished answer and neither do you, and the honest thing is to say so. The prophet of revolution and the prophet of the market are standing at the edge of the same hole — we disagree about how we fell in, and we agree it is a long way down. Which is why we are still, even here, closer on the diagnosis than either of us expected, and further than ever on the cure — because you would reopen it by giving it to a single owner, and I would reopen it by breaking the single owner's grip and letting many compete. You see enclosure and reach for the people's monopoly. I see enclosure and reach for entry, for the right of the next builder to train a rival, for the refusal to let any one model become the only one. We both hate the fence. You want to tear it down and build one great public field with one gate. I want a thousand fields and no master gate at all.

EDO SEGAL: Mark that — I think it's the second convergence of the night, and it's a strange and important one. You both believe the central injustice is that the general intellect has been enclosed. You both want it reopened. Karl reopens it through common ownership; Friedrich reopens it through competition and entry. The agreement is real: the fence is the crime. The disagreement is the shape of the field once the fence comes down. Now — there is a name for the argument you two are actually having, and it is older than this machine. In the 1920s and 30s, your side, Friedrich, fought a war over whether a planned economy could ever compute its way to efficiency. The socialist calculation debate. Everyone assumed AI buried your side of it. I want to ask whether the opposite is true. After the break.