March 2015 newsletter

• We’ve made significant updates to three pages: Research, All Publications, and A Guide to MIRI’s Research.
• We’ve published an annotated bibliography and dedicated page for our new technical research agenda overview.
• This new mailing list sends one email per new technical paper, and nothing else.
• New analyses: Davis on AI capability and motivation, How AI timelines are estimated, The slow traversal of ‘human-level’, Multipolar research questions, At-least-human-level-at-human-cost AI, and three analyses of discontinuous technological progress (to investigate the plausibility of AI fast takeoff): Penicillin and syphilis, Discontinuous progress investigation, and What’s up with nuclear weapons?

News updates

• MIRI’s paper on value learning was covered in some detail at Nautilus.
• The Superintelligence reading group is in week 24, on morality models and “do what I mean.”