One Decade of Universal Artificial Intelligence

The dream of creating artificial devices that reach or outperform human intelligence is an old one. A top-down approach is to first find a mathematical “gold-standard” definition of general intelligence, and then approximate the equations to make them computationally efficient. Indeed, Universal AI is such an elegant formal, objective, and non-anthropocentric theory. It results in an optimal reinforcement learning agent embedded in an arbitrary unknown environment that possesses essentially all aspects of rational intelligence. For the first time, without providing any domain knowledge, the same agent is able to self-adapt to a diverse range of interactive environments. These achievements give new hope that the grand goal of Artificial General Intelligence is not elusive. This tutorial provides an informal overview of UAI in context. It attempts to gently introduce a very theoretical, formal, and mathematical subject, and discusses philosophical and technical ingredients, traits of intelligence, underlying assumptions, relation to other work, differences to human intelligence, some social questions, and the past and future of UAI. Universal Artificial Intelligence 3 Marcus Hutter