When AI Levels the Playing Field: Skill Homogenization, Asset Concentration, and Two Regimes of Inequality

A new research paper by Xupeng Chen and Shuchen Meng, published on arXiv, presents a sophisticated economic model analyzing generative AI's paradoxical impact on inequality. The core finding reveals that while AI tools like GPT-4 or Claude 3 can homogenize skill levels within specific tasks—making a junior and senior writer produce similar-quality drafts—they simultaneously shift economic value away from labor and toward concentrated complementary assets. These assets include proprietary AI models, specialized datasets, and the capital required to deploy AI at scale. The model suggests this dynamic could create a society where individual performance gaps narrow, but wealth inequality widens as returns accrue to asset owners rather than workers.

The researchers formalize this tension using a task-based model with endogenous education choices and employer screening processes. Their analysis identifies two distinct economic regimes whose boundary depends critically on two factors: whether AI technology remains proprietary (like OpenAI's models) or becomes a commodity (like open-source Llama 3), and the structure of labor market institutions, particularly rent-sharing elasticity. Using a Method of Simulated Moments calibrated to six empirical targets, they show the aggregate sign of inequality change is highly sensitive to specific parameters. Notably, they demonstrate why standard occupational data (like BLS OEWS) cannot test their task-level predictions, calling for new within-occupation panel data that doesn't yet exist at scale.

This research moves beyond simplistic "jobs lost" debates to provide a mechanistic framework for understanding AI's distributional consequences. It highlights that the ultimate impact on inequality isn't predetermined by the technology itself, but will be shaped by policy choices around AI access, intellectual property, and labor market structures. The paper contributes a crucial analytical tool for policymakers and economists grappling with AI's second-order effects on the economy.