The Sufficient Machine Why the Impossibility of Superintelligence Does Not Prevent the Concentration of Power

A preceding study (Sophia, 2026) established four independent proofs that no currentAI architecture can produce superintelligence. This paper demonstrates that thisimpossibility is irrelevant to the actual consequences of AI deployment. The demanddriving hundreds of billions of dollars in AI investment is not superintelligence but twoconcrete objectives: workers who require no wages and soldiers who require nocourage. The technical threshold for autonomous weapons — perceptual patternmatching and real-time control — was crossed with deep learning circa 2012–2015.The threshold for large-scale cognitive labor replacement — human-averageperformance at low marginal cost — was crossed with large language models circa2022–2023. The present investment explosion coincides with the convergence of boththresholds, not with any proximity to superintelligence. The majority ceiling thatprevents superintelligence simultaneously provides the capability level sufficient forlabor displacement. Historical analysis of documented Cold War mind-controlprograms (MKUltra, 1953–1964) reveals that the institutional demand for perfectlycontrollable agents has persisted for seven decades; the method has shifted frommodifying human minds to replacing them with machines that never possessed theproperties one would need to suppress. The representation barrier that preventssuperintelligence also guarantees systematic harm in both military targeting andautomated services, concentrated among those least represented in training data. Thesuperintelligence debate — in both its utopian and doomer forms — functions asideological misdirection, directing attention toward a threshold that cannot be crossedand away from consequences already unfolding. The wall is permanent. The machineis sufficient. The rest is politics.