Marcus Hutter (born 14 April 1967 or 1 January 1967 in Munich) is a German and Australian computer scientist, professor, and artificial intelligence researcher. As a senior researcher at DeepMind (Google DeepMind), he investigates the mathematical foundations of artificial general intelligence.^[5,6] He is a professor at the Australian National University (in the College of Engineering, Computing and Cybernetics / ANU College of Engineering and Computer Science) in Canberra, Australia, from which he has been on leave.^[8]

Hutter studied physics and computer science at the Technical University of Munich and the Ludwig Maximilian University of Munich. Under the supervision of Harald Fritzsch and Wilfried Brauer, he completed his 1996 thesis "Instantons in QCD". In 2000, he joined Jürgen Schmidhuber's group at the Dalle Molle Institute for Artificial Intelligence Research (IDSIA / Istituto Dalle Molle di Studi sull'Intelligenza Artificiale) in Manno, Switzerland, and has also been associated with BrainLAB.^[2]

His primary research fields include artificial intelligence, Bayesian statistics, information theory, particle physics, and universal artificial intelligence. He developed a mathematical theory and formalism of artificial general intelligence named AIXI, and his book Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability was published by Springer in 2005.^[1] His notable doctoral students include Shane Legg, Jan Leike, and Tor Lattimore. Hutter has received numerous Best Paper Prizes, including Lindley 2006, Kurzweil AGI 2009, IJCAI-JAIR 2014, UAI 2016, AGI 2016, Alignment 2018, and IJCAI 2023, and maintains the website www.hutter1.net.

Marcus Hutter studied computer science and physics at the Technical University of Munich (TU München). In 1996, he received his doctorate under Harald Fritzsch on instantons in quantum chromodynamics. In 2000, he joined Jürgen Schmidhuber's group at the Swiss AI laboratory IDSIA (Dalle Molle Institute for Artificial Intelligence Studies) in Manno, Switzerland.^[6] There, he developed AIXI, a mathematical theory and formalism for optimal Universal Intelligence and artificial general intelligence based on Kolmogorov complexity and Ray Solomonoff's theory of universal inductive inference. In 2006, he accepted an appointment to a chair and worked as a professor at the Australian National University in Canberra.^[18] Since that same year, the Hutter Prize, named after him, has been awarded.

Starting in 2000 according to English and French sources, or in 2002 together with Jürgen Schmidhuber and Shane Legg according to Arabic and Spanish sources, Marcus Hutter developed and published a mathematical theory of artificial general intelligence, AIXI, based on idealized intelligent agents and reward-motivated reinforcement learning.^[9,1,7,8] His first book, Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability, was published in 2005 by Springer. Also in 2005, Hutter and his doctoral student Shane Legg published an intelligence test for artificial intelligence devices and systems.^[4,10,30] In 2009, Hutter developed and published the theory of feature reinforcement learning.^[11] In 2014, Lattimore and Hutter published an asymptotically optimal extension of the AIXI agent.^[12]

In 2019, Hutter joined DeepMind after being recruited by Shane Legg, and in 2022 he co-authored an article.^[7,31] An accessible podcast with Lex Fridman about his theory of Universal AI appeared in 2021, followed by a more technical follow-up with Tim Nguyen in the Cartesian Cafe in 2024.^[14,13] His new 2024 book provides a more accessible introduction to Universal AI and progress in the 20 years since his first book, including a chapter on ASI safety, which featured as a keynote at the inaugural workshop on AI safety in Sydney.^[16,15]

In 2006, Hutter announced the Hutter Prize for Lossless Compression of Human Knowledge, offering a total of €50,000 in prize money.^[17,1] In 2020, Hutter raised the prize money for the Hutter Prize to €500,000.^[18,1,17]

In 2002, Marcus Hutter published "The Fastest and Shortest Algorithm for All Well-Defined Problems" in the International Journal of Foundations of Computer Science. Springer published his book "Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability" in 2005. In 2007, Shane Legg and Marcus Hutter published "Universal Intelligence: A Definition of Machine Intelligence" in Minds and Machines. Hutter authored "A Complete Theory of Everything (will be subjective)", which appeared in the journal Algorithms in 2010. In 2011, Joel Veness, Kee Siong Ng, Marcus Hutter, William Uther, and David Silver published "A Monte-Carlo AIXI Approximation" in the Journal of Artificial Intelligence Research. Also in 2011, Samuel Rathmanner and Marcus Hutter published "A Philosophical Treatise of Universal Induction" in Entropy. Hutter published "Can Intelligence Explode?" in the Journal of Consciousness Studies in 2012. In 2015, Peter Sunehag and Marcus Hutter published "Rationality, Optimism and Guarantees in General Reinforcement Learning" in the Journal of Machine Learning Research. Reinhard Hutter and Marcus Hutter co-authored "Chances and Risks of Artificial Intelligence – A Concept of Developing and Exploiting Machine Intelligence for Future Societies", published in Applied System Innovation in 2021. In 2024, Marcus Hutter, David Quarel, and Elliot Catt authored "An Introduction to Universal Artificial Intelligence", published by Taylor & Francis.