Foom

Also: intelligence explosion, hard takeoff, fast takeoff, the singularity

The idea that once AI can improve itself, it could go from human-level to far beyond us very quickly.

“Foom” is the sound of something taking off very fast. It became the name for a fast “intelligence explosion” in a 2008 online debate between the AI theorist Eliezer Yudkowsky and the economist Robin Hanson, about “whether generally intelligent AIs will be able to improve their own capabilities very quickly”. The underlying idea is older: in 1965 the statistician I. J. Good argued that a machine smarter than us could design even better machines, setting off an “intelligence explosion”.

The mechanism is a feedback loop. If AI becomes good at the work of building AI, each improvement makes the next one come faster. The AI Safety Atlas uses “foom” interchangeably with “fast takeoff” and “hard takeoff”. The opposite view is a “slow takeoff”, in which progress still compounds, but more gradually; by one estimate the Atlas cites, still 10 to 100 times faster than the Industrial Revolution.

Which one we get matters more than it sounds. As the Atlas notes, the disagreement often comes down to whether problems would be “noticeable and fixable in time”. A fast takeoff leaves little room to spot a mistake and correct it. Scenarios such as AI 2027 put AI doing AI research at the centre of their story, and the EU’s Code of Practice for the most capable models lists AI “self-improvement” among the ways humans could lose control.