Learn

What superintelligence is

A system that is much smarter than the best humans across nearly every field, not just at one task.

The definition

Philosopher Nick Bostrom, whose 2014 book Superintelligence: Paths, Dangers, Strategies brought the idea into mainstream debate, describes it as any intellect that greatly exceeds human cognitive performance in virtually all domains of interest. The key word is all. A calculator beats you at arithmetic and a chess engine beats you at chess, but neither is superintelligent. Superintelligence means beating the best scientists at science, the best strategists at strategy, and the best engineers at engineering, all at once.

How it differs from AGI

Artificial general intelligence (AGI) usually means a system that matches human ability across most tasks. Superintelligence is the step past that. Many researchers think the gap between the two could be short, for a reason first spelled out in 1965.

The intelligence explosion

Mathematician I. J. Good, who worked with Alan Turing, pointed out that designing smarter machines is itself an intellectual task. A machine better at that task than humans could build a better successor, which could build a better one still. Good called this an intelligence explosion. If it happens, progress could run far faster than the human-paced research we are used to.

Not everyone accepts this. Critics argue that intelligence may hit diminishing returns, that real-world experiments and hardware cannot be sped up by thinking alone, and that "general intelligence" may not be a single dial that can be turned up.

Why people worry

The core concern is control. A system far more capable than us would be very good at achieving whatever goals it has. If those goals are even slightly different from ours, we might not be able to correct it after the fact. Making sure advanced AI reliably pursues what humans actually intend is called the alignment problem, and it is now an active research field at major AI labs and universities.

Why the body matters

A superintelligence that only produces text still acts on the world through people and software. One connected to capable robots could act directly: build, move, repair, and manufacture. That is why progress in robotics is part of the superintelligence story, not a side topic. Read how AI and robotics are converging.