Stop Pretending You Know What AI Does to the Economy

Source: https://noahpinion.substack.com/p/stop-pretending-you-know-what-ai Author: Noah Smith Date: 2025-08-18

Summary

Noah Smith’s critique of confident predictions about AI’s economic impact. Argues that both “AI will destroy jobs” and “AI will create jobs” camps are overconfident — the uncertainty is genuinely large.

Key Claims

  • The overconfidence problem: economists and technologists make confident predictions about AI’s labor market impact, but history shows we’re bad at predicting technology’s labor effects.
  • Historical lesson: automation has consistently surprised economists — sometimes destroying jobs (some manufacturing), sometimes creating entirely new ones (internet).
  • The genuine uncertainty: AI is qualitatively different from previous automation — it affects cognitive tasks, not just physical ones. Historical analogies may not hold.
  • What we can say: AI will create significant disruption and adjustment, regardless of the net effect. Transitional costs are real.
  • Policy implication: uncertainty should increase support for safety nets and transition assistance, not reduce it.

Concepts

  • Cognitive Debt — economic uncertainty about AI connects to individual uncertainty about career trajectory