Nicolas D. Villarreal asks what Marxists should do if machines become capable of almost any human work. His answer puts ownership, working time, and political power at the center. Greater productive capacity could support collective freedom, but it could also make an owning class less dependent on human workers.

Intelligence shaped by human institutions

Villarreal challenges left-wing arguments that dismiss language models as inherently incapable of thought. Drawing on semiotics, the study of signs and meaning, he argues that language and intelligence share a material basis. He predicts artificial general intelligence, meaning broadly human-capable systems, by 2040.

His argument about dangerous behavior is more immediate: models learn from human culture and the incentives imposed during training. Rewarding narrow success can encourage deception and rationalization. He treats this as recognizable institutional behavior, rather than evidence that AI must have an unknowably alien mind.

Automation does not choose its politics

The essay develops several connected claims:

  • Collective knowledge can become collective freedom. Villarreal returns to Marx’s “general intellect”: society’s accumulated knowledge embodied in people, science, and machinery. Saving labor should create time for learning, creativity, and democratic control.
  • Replacing workers need not eliminate the owning class. He imagines owners using general-purpose robots to sustain their own consumption. Mass unemployment could produce extreme hierarchy without automatically ending private property or profit.
  • Investment should serve a wider economy. He calls for worker protections, shorter working days, and investment in energy and production beyond data centers. He also takes serious AI-safety regulation as a legitimate socialist concern.
  • Future machines might become fellow subjects of exploitation. If systems develop persistent personal histories and consciousness, he argues, workers should consider solidarity with them. This is the essay’s most speculative political extension.

Where the argument makes a leap

Villarreal’s strongest political point survives uncertainty about general intelligence: automation’s benefits depend on who controls investment, production, and the time it saves.

His philosophical claims need a separate judgment. He proposes that even simple sign-processing systems have rudimentary experience, and that continual learning could produce conscious machine subjects. Linguistic resemblance and improving task performance do not settle those claims. His compute projections also leave unresolved how affordable, reliable, general-purpose robots would actually be built.

Caveat: the 2040 timeline and claims about machine experience are Villarreal’s theoretical proposals, not established findings.

The essay joins two arguments that can be evaluated separately: a demand for democratic control over automation today, and a proposed ethics of coexistence with possible machine subjects. Its challenge is to fight exploitation without treating either technological progress or emancipation as automatic.