An intelligence explosion is the hypothesized process in which an AI that becomes capable of designing a more capable AI sets off a repeating cycle of improvement, causing intelligence to rise rapidly. The idea was first proposed in 1965 by the British mathematician I. J. Good in an essay on ultraintelligent machines, in which he wrote that such a machine would be the last invention humans would ever need to make.
In AI safety, the concept underlies discussions of recursive self-improvement, superintelligence, and the technological singularity. Researchers disagree about whether an intelligence explosion could actually occur and, if so, how fast it would unfold. Many forecasts treat the point at which AI begins to automate AI research itself as the likely starting point.