Will There Be Only One Superintelligence?

July 4, 2026 – Max Rumpf

From a recent discussion with a friend:

Suppose some superintelligence gains a lead over its rivals. It uses that lead to improve itself faster than they can, and to accumulate resources by out-predicting both the world and the lesser intelligences in it. The loop compounds until it alone controls everything.

This story assumes the leader can predict the future well enough to pick the right research directions and investments. But superintelligence is itself the largest and most irreducible source of uncertainty about the future, and a lesser intelligence cannot predict a greater one. Its own descendants will be smarter than it is now, so it cannot know what they will do or which strategy is optimal for producing them. Its rivals' future versions will also be smarter than it is now, so it cannot predict them either.

Uncertainty doesn't make it impossible that the first superintelligence can maintain its lead, but it's not a foregone conclusion.