Universal Intelligence: A Definition of Machine Intelligence

By Shane Legg, Marcus Hutter. 2007

761 citations recorded by the source corpus.

A fundamental problem in artificial intelligence is that nobody really knows what intelligence is. The problem is especially acute when we need to consider artificial systems which are significantly different to humans. In this paper we approach this problem in the following way: we take a number of well known informal definitions of human intelligence that have been given by experts, and extract their essential features. These are then mathematically formalised to produce a general measure of intelligence for arbitrary machines. We believe that this equation formally captures the concept of machine intelligence in the broadest reasonable sense. We then show how this formal definition is related to the theory of universal optimal learning agents. Finally, we survey the many other tests and definitions of intelligence that have been proposed for machines.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Evaluation & Benchmarks

Citations in the corpus

Scalable agent alignment via reward modeling: a research direction (2018, background): “We desire alignment techniques that continue to work in the long term, i.e. that can scale to agents with superhuman performance in a wide variety of general domains (Legg & Hutter, 2007).”
Towards Deep Symbolic Reinforcement Learning (2016, background): “DRL can be thought of as a step towards instantiating the formal characterisation of universal artiﬁcial intelligence presented by Hutter [7], a theoretical framework for artiﬁcial general intelligence (AGI) founded on reinforcement learning.”
Overcoming catastrophic forgetting with hard attention to the task (2018, background): “Catastrophic forgetting is an important step in the advancement towards more general artiﬁcial intelligence systems (Legg & Hutter, 2007).”
On Learning the Geodesic Path for Incremental Learning (2021, background): “In pursuing a similar capability at an algorithmic level and in the so-called Incremental Learning (IL), an artificial machine must learn and remember various tasks or concepts [22] to accomplish a more broadened set of objectives.”
Acme: A Research Framework for Distributed Reinforcement Learning (2020, background): “A central goal of much of this work is to create a general agent that can learn to achieve goals across a wide range of environments (Legg and Hutter, 2007).”
DeepMind Lab (2016, background): “General intelligence measures an agent’s ability to achieve goals in a wide range of environments (Legg and Hutter, 2007).”
Multi-task Deep Reinforcement Learning with PopArt (2018, background): “Being able to acquire knowledge about a wide range of facts and skills has been long considered an essential feature for an RL agent to demonstrate intelligent behaviour (Sutton et al. 2011; Degris and Modayil 2012; Legg and Hutter 2007).”
The Measure of Intelligence (2019, neutral): “We thus disagree with the perspective of Universal Psychometrics [39] or Legg and Hutter’s Universal Intelligence [54], which reject anthropocentrism altogether and seek to measure all intelligence against a single absolute scale.”
Evaluation in artificial intelligence: from task-oriented to ability-oriented measurement (2017, background): “…and Schultz, 1988; Decker et al, 1989; Langley, 1987, 2011; Buchanan, 1988; Simon, 1995; Baldwin and Yadav, 1995; Falkenauer, 1998; Langford, 2005; Legg and Hutter, 2007a; Whiteson et al, 2011; Drummond and Japkowicz, 2010; Anderson et al, 2011; Madhavan et al, 2009; Schlenoff et al, 2011) are…”
Aligning Superintelligence with Human Interests: A Technical Research Agenda (2015, background): “Indeed, the interaction problem (and AIXI [Hutter 2000], its solution) are the basis for the “universal measure of intelligence” developed by Legg and Hutter (2007).”
Agent Foundations for Aligning Machine Intelligence with Human Interests: A Technical Research Agenda (2017, background): “Consider a counterfactual in which a given deterministic decision procedure selects a different action from the one it selects in reality.”
Avoiding Wireheading with Value Reinforcement Learning (2016, background): “As Bostrom (2014b) convincingly argues, it is important that we find a way to specify robust goals for superintelligent agents.”