AGI-16 Tom Everitt - Controlling Arbitrarily Intelligent Systems

Tom Everitt presents his talks “Self-Modification of Policy and Utility Function in Rational Agents“ and “Avoiding Wireheading with Value Reinforcement Learning“ (starting around 15:30) at the Ninth Conference on Artificial General Intelligence (AGI-16) in New York (). This paper won the Kurzweil Prize for Best AGI Paper. Self-Modification of Policy and Utility Function in Rational Agents (winner of the Kurzweil Prize for Best AGI Paper) Authors: Tom Everitt, Daniel Filan, Mayank D