A Simple Model of How AI Could Shape Economic Growth

Summary

Tyler Cowen argues that the standard Solow growth model misses an important feature of rapid AI-driven change. He proposes treating production as the combination of two factors: “intelligence,” such as formal reasoning and mathematical ability, and “Polanyi knowledge,” the tacit, context-specific understanding of how people and organizations actually work. AI is rapidly increasing the supply of the first factor, while humans remain especially important to the second; AI may assist with contextual work, but cannot simply be placed in an office and expected to understand its norms and routines. Because the factors are mostly complements rather than easy substitutes, more AI intelligence could raise the marginal value, employment, and real wages of workers who provide contextual knowledge. Cowen expects some transitional unemployment in intelligence-focused work as human-AI “Centaur” arrangements fade, although he notes that such arrangements are still supporting work such as mathematics. He expects gains in the Polanyi-knowledge sector to arrive slowly because that knowledge is messy and cannot be expanded quickly or directly, but to continue over a long period as organizations absorb and exploit AI advances. The framework also suggests that control of the intelligence sector alone would confer less power over society than it might seem, because other necessary complements remain scarce. Cowen says the model draws on questions he considered in his youth about Soviet cybernetics and central planning. He presents it as a starting point, and argues that it broadly fits current conditions: major technology advances alongside a functioning job market and robust, but not explosive, economic growth. He expects growth could rise as the contextual-knowledge sector gradually catches up.