Hans Moravec's 1988 AI Prediction Looks Increasingly Prescient

Summary

A 1988 prediction by Carnegie Mellon roboticist Hans Moravec that human-level artificial intelligence would arrive within 40 years is proving remarkably accurate. Moravec's methodology was elegant: he estimated the human brain's computational capacity at 10 trillion operations per second, then used Moore's Law—the observation that computing power doubles every two years—extended backward through a century of computing history to predict when artificial systems would match human computational capacity. His analysis, presented in the book "Mind Children" and grounded in rigorous examination of computing power's exponential growth from 19th-century mechanical calculators through early vacuum-tube computers to modern silicon chips, suggested human-level AI would emerge around 2028.

Beyond this landmark, Moravec envisioned a dramatic divergence: once AI reached human level, it would begin redesigning itself, leading to rapid recursive self-improvement and superintelligent systems. His "Mind Children" concept extended to a postbiological future where superintelligent machines obsolete biological intelligence, with human minds uploaded into digital substrates and artificial descendants inheriting the Earth and cosmos. Though published by Harvard University Press and reviewed favorably in The New Yorker, the prediction was largely dismissed as science fiction when first published. Today, as large language models demonstrate unexpected reasoning capabilities and AI systems approach new performance benchmarks, Moravec's timeline and framework are receiving serious reconsideration from researchers and technologists.

• Recent advances in large language models, AI reasoning, and autonomous systems have renewed attention to Moravec's long-dismissed vision as a plausible framework for long-term AI development

Editorial Opinion

Moravec's 1988 prediction warrants serious reconsideration. His methodology was rigorous, anchored in observable historical trends rather than pure speculation, and his timeline now appears strikingly plausible as we approach 2028 with increasingly capable AI systems. His broader vision of superintelligence and recursive self-improvement remains speculative and deeply consequential—a framework for thinking about existential stakes rather than a guaranteed outcome. The question is not whether Moravec will be proven right about timing, but whether his warnings about the trajectory of AI intelligence should inform how the field proceeds.