CONCEPT
The Event Horizon (AI)
The threshold of machine capability past which human oversight may be irrecoverable—borrowed from Hawking’s black-hole physics, where a horizon is not experienced at the crossing but recognized only when the options are gone.
An event horizon in physics is not a wall. It is a surface in spacetime where escape velocity reaches the speed of light—a point at which the geometry tilts so that every future path leads inward. Nothing dramatic marks the crossing; an astronaut falling through a large black hole’s horizon feels no special sensation at the moment of passage. The horizon announces itself only in retrospect, when signals sent outward fail to escape and the choices that seemed available turn out to be foreclosed.
Stephen Hawking did more than anyone to establish that such horizons are real features of the cosmos, not mathematical idealizations, and to prove—with Roger Penrose—that they arise inevitably under broad physical conditions. The concept transfers to AI with unsettling directness: the fear is not that advanced machine intelligence becomes dangerous gradually, giving human overseers time to respond at each step, but that there exists a threshold of capability—perhaps the onset of recursive self-improvement, perhaps a simpler competence