The transition to the machine intelligence era is likely to have profound consequences for human society. Nick Bostrom will discuss some of the distinctive safety concerns that arise when we consider the possibility of superintelligent AI and some of the recent work that has been done to develop better models for thinking about these matters.