Marcus Hutter is an AI researcher known for trying to give general intelligence a precise mathematical definition. In his 2005 book “Universal Artificial Intelligence” he introduced AIXI, a theoretical agent that, given unlimited computation, would behave optimally in any computable environment. AIXI combines sequential decision theory with Ray Solomonoff’s theory of universal induction: it considers all possible programs that could explain its observations, weights simpler explanations more heavily, and chooses actions to maximize expected reward. It is uncomputable in practice - a mathematical ideal rather than a runnable algorithm - but it serves as a reference point for what “optimal” general intelligence would even mean.
With his student Shane Legg, Hutter co-authored the 2007 paper “Universal Intelligence: A Definition of Machine Intelligence,” which turned this machinery into a single formal measure of an agent’s intelligence across all environments. The pair distilled common features from many informal expert definitions of intelligence and expressed them as one equation, an approach that influenced later attempts to define and benchmark generality.
Hutter has held positions at IDSIA in Switzerland, the Australian National University, and Google DeepMind. He also funds the Hutter Prize, a cash award for compressing a large sample of human knowledge, built on his thesis that better compression of text is a direct proxy for better understanding and therefore for intelligence.