AI's Diminishing Returns: Why Superintelligence May Not Transform the World

AI has already achieved superhuman performance in mathematics, disproving the Jacobian Conjecture and solving quantum cryptography problems, yet the world remains largely unchanged. Productivity growth has been modest, and employment has not been disrupted. Researchers like Francois Chollet argue that intelligence is subject to diminishing returns—like making a ball rounder rather than a tower taller—due to limited data and irreducible error in complex systems. Arvind Narayanan and Sayash Kapoor predict AI will not outperform trained humans at forecasting geopolitical events. However, AI's replicability, ability to synthesize distributed tacit knowledge (e.g., Zeiss glass manufacturing), and capacity to discover 'cloud laws' in complex phenomena could still drive productivity gains. The key advantage is not raw intelligence but AI's computer-like ability to handle vast data and interface with sensors, potentially enabling a new industrial revolution and scientific renaissance.