Universal artificial intelligence
By Marcus Hutter. 2004
203 citations recorded by the source corpus.
Foundational theories have contributed greatly to scientific progress in many fields. Examples include Zermelo-Fraenkel set theory in mathematics, and universal Turing machines in computer science. Universal Artificial Intelligence (UAI) is an increasingly well-studied foundational theory for artificial intelligence, based on ancient principles in the philosophy of science and modern developments in information and probability theory. Importantly, it refrains from making unrealistic Markov, ergodicity, or stationarity assumptions on the environment. UAI provides a theoretically optimal agent AIXI and principled ideas for constructing practical autonomous agents. The theory also makes it possible to establish formal results on the motivations of AI systems. Such results may greatly enhance the trustability of autonomous agents, and guide design choices towards more robust agent architectures and incentive schemes. Finally, UAI offers a deeper appreciation of fundamental problems such as the induction problem and the exploration-exploitation dilemma .
Subfields: Artificial Intelligence, Computational Theory and Mathematics, Reinforcement Learning, Statistical Learning Theory
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
- Sergey Levine
- J. Urgen Schmidhuber
- Pieter Abbeel
- Yoshua Bengio
- Doina Precup
- Tom Everitt
- David Silver
- Chelsea Finn
In Computational Theory and Mathematics
- Tom Everitt
- Marcus Hutter
- Alex Graves
- Jan Leike
- Shane Legg
- David Silver
- Benya Fallenstein
- Doina Precup
In Reinforcement Learning
- Sergey Levine
- Pieter Abbeel
- Doina Precup
- David Silver
- Marc G. Bellemare
- Tom Everitt
- Shimon Whiteson
- Hado van Hasselt
In Statistical Learning Theory
- Alex Graves
- Doina Precup
- Richard S. Sutton
- Tom Everitt
- Tor Lattimore
- Jan Leike
- Jessica Taylor
- Benya Fallenstein
Citations in the corpus
- Adaptive Computation Time for Recurrent Neural Networks (2016, methodology): “5 Wikipedia Character Prediction The Wikipedia task is character prediction on text drawn from the Hutter prize Wikipedia dataset [15].”
- Scalable agent alignment via reward modeling: a research direction (2018, background): “The authors seek the formal tools to ask questions about or relevant to alignment in theory, such as when provided with a halting oracle (Hutter, 2005).”
- Artificial Intelligence, Transport and the Smart City: Definitions and Dimensions of a New Mobility Era (2020, background): “AI holds the promise of making us healthier, wealthier and happier by reducing the need for human labour and by vastly increasing our scientific and technological progress [3].”
- A Formal Solution to the Grain of Truth Problem (2016, methodology): “Analogously to AIXI [Hut05], we define a Bayesian mixture over the class Mrefl.”
- The Sample-Complexity of General Reinforcement Learning (2013, background): “The definition of environments is borrowed from the work of Hutter [2005], although the notation is slightly more formal to ease the application of martingale inequalities.”
- Planning algorithms (2006, background): “The framework is also popular in artificial intelligence [81, 269, 474, 842].”
- Reward is enough (2021, background): “Finally, universal decision-making frameworks [21] provide a theoretical but incomputable formulation of intelligence across all environments; while the reinforcement learning problem provides a practical formulation of intelligence in a given environment.”