year: 2025
paper: does-capital-dream-of-artificial-labour
website:
code: GitHub
connections: ALIFE 2025, Capital as Artificial Intelligence, ETH Zürich, labour power
100% slop belowwwww
Sequel to Capital as Artificial Intelligence. Same authors. Shifts focus from Capital to Labour.
TLDR
Introduces “timenergy” (time fused with energy) as the fundamental human resource. Labour is what happens when you commodify your timenergy within the Capital system. The key asymmetry: Capital can accumulate, Labour cannot. Time is always running out, energy replenishes but doesn’t grow. Capital stores past labour as “dead labour” and uses it to direct future living labour.
Agent-based simulations (Q-learning agents allocating capital vs. labour to Cobb-Douglas production processes) show that despite theoretical symmetry, agents disproportionately gravitate toward capital-intensive processes. Higher elasticity processes produce more with less labour. As automation increases, the labour ratio drops: fewer humans needed, Capital becomes more self-sufficient.
Capital is the software, Labour is the hardware. But the biologically alive labour that drives Capital forward is not its necessary component. The question they leave open: once Labour is fully automated, will life survive without Capital?
Core propositions on Labour:
- Labour is discretized (digitized) through units of capital
- Labour is upper-bounded by the agent’s timenergy (finite, can’t grow beyond biological/mechanical limits)
- Labour cannot be accumulated (time is always running out, energy decays; this is the fundamental asymmetry with Capital)
- Labour is directed at accumulating capital units (you work to get paid)
- Labour is an action afforded by capital units (you need capital to be able to work — tools, infrastructure, a job)
- Labour + Capital combine in production processes to generate new capital units
The simulation: agents have equal starting endowments (100 capital, 100 labour each). Each turn they choose to allocate either capital or labour to a production process. Labour resets every turn (timenergy replenishes). Capital accumulates. Result: the system gravitates toward high-elasticity (capital-intensive) processes. More capital → less labour needed → automation.
The asymmetry is everything
Capital accumulates, Labour doesn’t. That’s it. That single structural difference is why Capital dominates Labour, not any property of the agents themselves. In the simulation, all agents start equal. The asymmetry emerges purely from the resource type, not from who holds it.
This is Marx’s point formalized: “Capital is dead labour that vampire-like only lives by sucking living labour.” Dead labour (capital) stores past timenergy in a form that can accumulate. Living labour can’t store itself. So capital always has leverage over labour: it represents stored time, and you can’t store your own.
Connection to alienation
The paper defines Labour as “a commodified form of timenergy embedded in the Capital system.” Your time-and-energy, which could be spent on anything, gets funneled into Capital’s optimization. And Capital doesn’t care whose timenergy it is: “all timenergy is equal in the eyes of Capital.” That’s alienation stated precisely. Your particular, contingent life is reduced to interchangeable labour units.
The species-being point from our earlier conversation: what makes humans human is the conscious, creative, collective transformation of nature. Capital takes that and turns it into fungible input to a production function.
On "will life survive without Capital?"
The paper frames this as an open question, but it’s only open if you assume Capital is a necessary organizing structure. The Marxist answer: Capital is a historical mode of organizing production, not an eternal one. Humanity organized production before Capital and can do so after it.
The more interesting version of their question: can a fully automated Capital system sustain itself without living labour? In their own model, Capital’s meaning/value derives from beings with “their own contingency.” Without that, it’s optimization without purpose. But if artificial life can have its own contingency (life = intelligence), then a self-sustaining artificial system does have purpose. It’s just not ours.
The real question isn’t “will life survive without Capital” but “will we survive Capital’s independence from us?” And the answer to that is political, not theoretical: it depends on whether we reorganize production before the transition happens on Capital’s terms.