The Future of Humanity Institute and DeepMind are co-hosting monthly seminars aimed at deepening the ongoing fruitful collaboration between AI safety researchers in these organisations. The Future of Humanity Institute played host to the seminar series for the first time last week.
DeepMind’s Laurent Orseau has been collaborating with Stuart Armstrong on “Safely Interruptible Agents“, among other topics. Orseau is also working with FHI’s Jan Leike on issues including “Thompson sampling is asymptotically optimal in general environments“, which won best student paper at UAI 2016.