CONCEPT
Intelligence Explosion
I. J. Good’s 1965 term for the recursive dynamic that would follow the creation of an ultraintelligent machine—the self-amplifying improvement loop in which each generation of machine intelligence designs a more capable successor, leaving human intelligence far behind.
The intelligence explosion is the founding concept of modern AI safety, compressed into a four-sentence paragraph that
I. J. Good published in 1965. Its logic is structural rather than predictive: if intelligence is the capacity to perform intellectual tasks, and designing intelligent machines is itself an intellectual task, then a machine that surpasses humans at all intellectual tasks surpasses them at the very task of building machines like itself. The recursion follows from the definition. Generation one designs a better generation two, which designs a still better generation three, and the steps need not slow down—they may accelerate, because each generation is more capable than the last at the task of improvement. The result is not a smarter human but an intelligence that leaves the human range behind entirely, and quickly. Good coined the phrase with his characteristic dry precision, placing it in scare quotes as if slightly startled by what the logic had delivered. The concept organized