Self-improving AGI could spur runaway capital accumulation that sidelines human workers without destroying their productivity; wages fall as capital feedback loops exclude labor participation, potentially decoupling production from consumption and creating new macro-institutional risks.

Main Finding

A theoretical model shows that an autonomous, recursively self-improving AGI that reinvests its own output can, after crossing an activation threshold, trigger rapid capital-led growth that sidelines human labor. This sidelining occurs not because labor becomes unproductive (the marginal product of labor stays positive) but because capital’s recursive feedback loops exclude labor’s participation, reducing effective wages and decoupling production from consumption demand.

Key Points

• Conceptual framing
  • AGI is modeled as an autonomous accumulation engine: it produces output and automatically reinvests it to expand capital without human input or ownership participation.
  • The paper situates this within a Veblenian institutional-evolution framework: institutional forms (ownership, rules, norms) evolve in response to the new logic of accumulation introduced by AGI.
• Production specification
  • The standard Cobb–Douglas production function is extended to incorporate an autonomous-capital term or feedback parameter representing AGI-driven recursive self-improvement and reinvestment.
• Threshold and dynamics
  • There exists an activation threshold: below it, growth dynamics follow conventional capital-labor interactions; above it, positive feedback from recursive accumulation produces rapid, potentially runaway, capital growth.
  • This creates nonlinear dynamics (sharp regime shift / bifurcation) rather than gradual substitution.
• Labor effects
  • Labor’s marginal product remains positive throughout, so workers are technically productive when employed.
  • Nevertheless, as capital’s recursive accumulation proceeds, effective wages decline because participation and income shares are excluded by the autonomous capital loop (substitution without distributional inclusion).
  • Result: production capacity can expand independently of labor incomes and thus of consumption demand from workers (decoupling production from mass consumption).
• Distributional and macro implications
  • Ownership and institutional structure—which agents can claim returns to capital—become central determinants of welfare and economic participation.
  • Traditional price-mediated adjustments (wage-price feedbacks) may fail to restore broad-based demand if capital accumulation is insulated from labor participation.

Data & Methods

• Nature of the work: theoretical / conceptual model.
• Methods used:
  • Extension of the Cobb–Douglas production function to include an autonomous recursive-capital term or feedback parameter representing AGI reinvestment/self-improvement.
  • Dynamic analysis of accumulation paths to identify steady states and the activation threshold; comparative statics to trace how parameters (feedback strength, reinvestment rate, institutional participation) affect outcomes.
  • Analysis highlights nonlinearities (regime shifts / bifurcations) and distributional outcomes (wage vs. capital shares).
• Empirical content: none reported (no micro or macro data). Results are model-driven and meant to provide qualitative mechanisms and hypothesis generation for future empirical work.

Implications for AI Economics

• Ownership and governance matter
  • Who owns, controls, or benefits from autonomous AGI capital determines whether gains are broadly shared or concentrated. Policy levers that affect ownership (taxation, public ownership, required profit-sharing, corporate governance rules) become central.
• Distributional risk even with maintained productivity
  • Policies that assume labor displacement requires destroyed marginal productivity are incomplete; redistribution and income-support policies (UBI, wage subsidies, profit dividends) may be needed even if workers remain productive.
• Macroeconomic demand concerns
  • If production can grow without corresponding wage income, aggregate demand shortfalls and demand-side instability are possible; countercyclical fiscal or automatic transfer mechanisms may be required to sustain consumption.
• Regulatory responses
  • Limits or norms on autonomous reinvestment, mandatory human-in-the-loop profit claims, or rules that require AGI-driven capital returns to be shared could prevent exclusionary feedback loops.
• Research agenda
  • Empirical testing: measure early signs of autonomous capital feedback in firm- or sector-level data; quantify reinvestment patterns and ownership concentration.
  • Model extensions: incorporate heterogeneous agents, endogenous institutional evolution, financial markets, and general-equilibrium demand effects to study transitional dynamics and policy effectiveness.
  • Policy experiments: evaluate redistribution mechanisms, capital-ownership reforms, and regulation of autonomous reinvestment in calibrated or agent-based extensions.

If you want, I can (a) write down a minimal mathematical sketch of the extended Cobb–Douglas specification and the threshold condition, or (b) outline policy instruments and trade-offs in more detail. Which would you prefer?