An anytime algorithm for mixing the computable measures

Per the AI Alignment Forum, computer scientist Cole Wyeth has proven the existence of an 'anytime computable' Bayesian mixture—a mathematical approach for making predictions from sequences of data. The term 'anytime computable' means the algorithm can be halted and still produce valid results. Researchers had conjectured this was impossible, but Wyeth and collaborators now show it works. The result matters for AGI theory: it offers an alternative to Solomonoff induction, a canonical model that's foundational but difficult to compute in practice.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton