AI systems excel where massive digital training data exist, but they lack Polanyi (tacit, context‑specific) knowledge embedded in physical practice and social relations. That gap means many tasks—especially those requiring nuanced hands‑on judgment or motivating people—are harder to automate than headline 'AGI' claims suggest. Policy and regulation debates that assume broad job displacement overlook this structural limitation.
— If tacit (Polanyi) knowledge is central to many occupations, then predictions about mass automation, labor policy, and AI regulation need to be recalibrated toward role‑specific risk and institutionally focused responses.
Arnold Kling
2026.10.16
90% relevant
Luis Garicano’s Smith‑vs‑Coase framing argues the balance between cheap knowledge (models) and cheap transactions will determine whether work is internalized or outsourced, and highlights that hard‑to‑automate, rare, verification‑heavy expertise (tacit knowledge) retains value—this is a concrete restatement of the existing idea that tacit work resists automation.
Arnold Kling
2026.09.14
100% relevant
Arnold Kling (citing Tyler Cowen) contrasts 'intelligence' (formal, digitizable skills) with 'Polanyi knowledge' (inarticulable, place‑bound know‑how) and ties the latter to the Solow residual and cross‑country productivity differences.
← Back to all ideas