On the Limits of Recursively Self-Improving AGI

Self-improving software has been a goal of computer scientists since the founding of the field of Artificial Intelligence. In this work we analyze limits on computation which might restrict recursive self-improvement. We also introduce Convergence Theory which aims to predict general behavior of RSI systems.

[1]  Ben Goertzel,et al.  Why an Intelligence Explosion is Probable , 2012 .

[2]  Jonathan Schaeffer,et al.  Checkers Is Solved , 2007, Science.

[3]  Luca Bertazzi,et al.  Reoptimizing the traveling salesman problem , 2003, Networks.

[4]  Eliezer,et al.  Tiling Agents for Self-Modifying AI , and the Löbian Obstacle * , 2013 .

[5]  S. Lloyd.  Ultimate physical limits to computation , 1999, Nature.

[6]  Douglas Fox,et al.  The limits of intelligence. , 2011, Scientific American.

[7]  Alexey V. Melkikh,et al.  The No Free Lunch Theorem and hypothesis of instinctive animal behavior , 2014, Artif. Intell. Res..

[8]  Benja Fallenstein,et al.  Problems of Self-reference in Self-improving Space-Time Embedded Intelligence , 2014, AGI.

[9]  Hans J. Bremermann,et al.  Quantum noise and information , 1967 .

[10]  Roman V. Yampolskiy,et al.  Analysis of Types of Self-Improving Software , 2015, AGI.

[11]  Miri Visiting Fellow,et al.  Coherent Extrapolated Volition: A Meta-Level Approach to Machine Ethics , 2010 .

[12]  A. Turing.  On Computable Numbers, with an Application to the Entscheidungsproblem. , 1937 .

[13]  Matt Mahoney.  A Model for Recursively Self Improving Programs , 2010 .

[14]  Roman V. Yampolskiy,et al.  The Universe of Minds , 2014, ArXiv.

[15]  CERN,et al.  Universal Limits on Computation , 2004 .

[16]  Dafna Shahaf,et al.  Towards a Theory of AI Completeness , 2007, AAAI Spring Symposium: Logical Formalizations of Commonsense Reasoning.

[17]  Juraj Hromkovic,et al.  On the Hardness of Reoptimization , 2008, SOFSEM.

[18]  Roman V. Yampolskiy,et al.  AI-Complete, AI-Hard, or AI-Easy - Classification of Problems in AI , 2012, MAICS.

[19]  D. Chalmers.  The Singularity: a Philosophical Analysis , 2010 .

[20]  Roman V. Yampolskiy,et al.  Artificial Intelligence Safety Engineering: Why Machine Ethics Is a Wrong Approach , 2011, PT-AI.

[21]  Roman V. Yampolskiy,et al.  Turing Test as a Defining Feature of AI-Completeness , 2013, Artificial Intelligence, Evolutionary Computing and Metaheuristics.

[22]  Eliezer Yudkowsky,et al.  Levels of Organization in General Intelligence , 2007, Artificial General Intelligence.

[23]  J. Bekenstein.  Information in the holographic universe. , 2003 .

[24]  Marcus Hutter,et al.  Can Intelligence Explode? , 2012, ArXiv.

[25]  Roman V. Yampolskiy,et al.  Utility function security in artificially intelligent agents , 2014, J. Exp. Theor. Artif. Intell..

[26]  Roman V. Yampolskiy,et al.  AI-Complete CAPTCHAs as Zero Knowledge Proofs of Access to an Artificially Intelligent System , 2012 .

[27]  A. Einstein.  Ist die Trägheit eines Körpers von seinem Energieinhalt abhängig? [AdP 18, 639 (1905)] , 2005, Annalen der Physik.

[28]  Sang Joon Kim,et al.  A Mathematical Theory of Communication , 2006 .

[29]  Jirí Wiedermann.  A Computability Argument Against Superintelligence , 2012, Cognitive Computation.

[30]  Scott Aaronson,et al.  Guest Column: NP-complete problems and physical reality , 2005, SIGA.

[31]  J. Storrs Hall,et al.  Engineering Utopia , 2008, AGI.

[32]  H. Rice.  Classes of recursively enumerable sets and their decision problems , 1953 .

[33]  Roman V. Yampolskiy.  Computing Partial Solutions to Difficult AI Problems , 2012, MAICS.

[34]  Richard Yonck,et al.  Toward a Standard Metric of Machine Intelligence , 2012 .

[35]  Stephen Wolfram,et al.  A New Kind of Science , 2003, Artificial Life.

[36]  David H. Wolpert,et al.  No free lunch theorems for optimization , 1997, IEEE Trans. Evol. Comput..

[37]  Thomas Bolander,et al.  Logical Theories for Agent Introspection , 2004 .

[38]  Vangelis Th. Paschos,et al.  Reoptimization of minimum and maximum traveling salesman's tours , 2009, J. Discrete Algorithms.

[39]  A. Kolmogorov.  Three approaches to the quantitative definition of information , 1968 .

[40]  J. Wheeler.  Information, physics, quantum: the search for links , 1999 .

[41]  Nick Bostrom,et al.  Superintelligence: Paths, Dangers, Strategies , 2014 .

[42]  Roman V. Yampolskiy.  Construction of an NP Problem with an Exponential Lower Bound , 2011, ArXiv.

[43]  A. Sandberg,et al.  The Physics of Information Processing Superobjects : Daily Life Among the Jupiter Brains , 1999 .

[44]  Laurent Orseau,et al.  Self-Modification and Mortality in Artificial Agents , 2011, AGI.

[45]  Roman V. Yampolskiy,et al.  Efficiency Theory : a Unifying Theory for Information, Computation and Intelligence , 2011, Journal of Discrete Mathematical Sciences and Cryptography.

[46]  Nick Bostrom.  What is a Singleton , 2006 .

[47]  Marcus Hutter,et al.  Universal Algorithmic Intelligence: A Mathematical Top→Down Approach , 2007, Artificial General Intelligence.

[48]  Albert Einstein,et al.  Does the Inertia of a Body Depend upon Its Energy-content? , 2001 .

[49]  G. Ausiello,et al.  Complexity and Approximation in Reoptimization , 2008 .