"AI safety and developmental interpretability
Daniel Murfet
More scientists should work part-time on AI safety. This talk describes several outstanding technical problems in AI safety and how they relate to existing disciplines, and outline a research agenda being pursued by a group of us at the University of Melbourne in collaboration with others around the world, which aims at making fundamental breakthroughs in interpretability of large neural networks.
Daniel Murfet is a mathematician at the University of Melbourne, with interests in algebraic geometry, logic and theory of deep learning. He was a signatory to an open letter to Ed Husic MP, calling for him to take AI safety seriously.
MC: Zac Broeren
Find out more about EA Global conferences at: www.eaglobal.org
Learn more about effective altruism at: www.effectivea...