April 2016 Newsletter

Research updates

• A new paper: “Parametric Bounded Löb’s Theorem and Robust Cooperation of Bounded Agents“
• New at IAFF: What Does it Mean for Correct Operation to Rely on Transfer Learning?; Virtual Models of Virtual AIs in Virtual Worlds

General updates

• We’re currently accepting applicants to two programs we’re running in June: our 2016 Summer Fellows program (details), and a new Colloquium Series on Robust and Beneficial AI (details).
• MIRI has a new second-in-command: Malo Bourgon.
• We’re hiring! Apply here for our new research position in type theory.
• AI Impacts is asking for examples of concrete tasks AI systems can’t yet achieve. You can also submit these tasks to Phil Tetlock, who is making the same request for Good Judgment Open.
• MIRI senior researcher Eliezer Yudkowsky discusses his core AI concerns with Bryan Caplan. (See Caplan’s response and Yudkowsky’s follow-up.)
• Yudkowsky surveys lessons from game-playing AI.

News and links

• Google DeepMind’s AlphaGo software defeats leading Go player Lee See-dol 4-1. GoGameGuru provides excellent commentary on each game (1, 2, 3, 4, 5). Lee’s home country of South Korea responds with an AI funding push.
• In other Google news: The New York Times reports on an AI platform war; Alphabet’s head of moonshots rejects AI risk concerns; and Alphabet jettisons its main robotics division.
• The UK Parliament is launching an inquiry into “social, legal, and ethical issues” raised by AI, and invites written submissions of relevant evidence and arguments.
• The White House’s Council of Economic Advisers predicts the widespread automation of low-paying jobs. Related: How Machines Destroy (And Create!) Jobs.
• CGP Grey, who discussed automation in Humans Need Not Apply (video), has a thoughtful conversation about Nick Bostrom’s Superintelligence (audio).
• Amitai and Oren Etzioni call for the development of guardian AI, “second-order AI software that will police AI.”
• In a new paper, Bostrom weighs the pros and cons of openness in AI.
• Bostrom argues for scalable AI control methods at RSA Conference (video).
• The Open Philanthropy Project, a collaboration between GiveWell and Good Ventures, awards $100,000 to the Future of Life Institute.
• The Center for Applied Rationality is seeking participants for two free programs: a Workshop on AI Safety Strategy and EuroSPARC, a math summer camp.