An intellectual analogy compares AI safety's Joe Carlsmith and Eliezer Yudkowsky to philosophers G.A. Cohen and Karl Marx

Sharp analogy. The through-line is: careful analytic philosopher who takes the core claims of a heterodox, polemical system-builder and reformulates them in a way the academy can engage with on its own terms.

Cohen did this explicitly with *Karl Marx's Theory of History* — taking historical materialism and defending it using the tools of mainstream analytic philosophy (the "No-Bullshit Marxism" crew). Carlsmith does something similar with AI x-risk: his OpenPhil reports read like someone translating Yudkowsky's intuitions about power-seeking AI into the kind of careful probabilistic reasoning and philosophical argumentation that doesn't require you to have read the Sequences.

One tension worth noting: both Cohen and Carlsmith end up being more interesting than the "translator" framing suggests. Cohen eventually developed positions (on self-ownership, on the ethos of justice) that cut against orthodox Marxism. Carlsmith's stuff on scheming and moral patienthood goes places Yudkowsky hasn't, or wouldn't. So the relationship is less "formalize the master" and more "take the generative core seriously enough to follow it somewhere the originator didn't."

Also an implicit claim in the analogy: that Yudkowsky occupies a Marx-like position as the foundational but academically marginal originator whose ideas are more influential than his methodology would normally permit. That feels roughly right for the AI safety world specifically.