What superintelligence means, who defined it, and who is arguing about it
Seven sourced guides to the word superintelligence: the original definitions, how AI companies use the term today, the forecasts, and the public debate. Each claim links to its source, and each date-sensitive fact is stated as of October 2026. Sources are listed at the bottom of every guide.
All guides
What is superintelligence?
Nick Bostrom defined superintelligence as an intellect that greatly exceeds human performance in virtually all domains. This guide traces the idea from I. J. Good in 1965 to the bills introduced in 2026.
AGI vs superintelligence
There is no agreed definition of AGI, and the competing ones measure different things. This guide compares OpenAI's economic test, Google DeepMind's Levels of AGI and Anthropic's powerful AI.
How AI companies define superintelligence
OpenAI, Meta, Safe Superintelligence Inc. and Microsoft name superintelligence as a goal, but mean different things by it. Anthropic and Google DeepMind mostly use other terms for similar capabilities.
Types of superintelligence
Nick Bostrom sorts superintelligence into speed, collective and quality forms. Other schemes sort it by path, by breadth, by the role it plays, or by how fast it could emerge.
The intelligence explosion and the singularity
I. J. Good argued in 1965 that a machine able to design better machines would set off an intelligence explosion. This guide sets out that argument, its later versions, and the main objections.
Superintelligence timelines
Forecasts for human-level or superhuman AI range from the late 2020s to the next century. This guide lists who said what and when, and explains why definitions drive much of the disagreement.
The superintelligence debate
Some researchers and lab leaders warn that superintelligence could escape human control; skeptics say it is distant, ill-defined or marketing. This guide sets out both sides and the policy responses through October 2026.