Artificial Intelligence and Control of Humanity’s Future

Abstract: 

Current machine learning models perform many cognitive tasks at or beyond human level; future AI may outstrip us in every meaningful way. Under these circumstances, we’ll face pressures to cede important decisions to AI systems. Continuing sufficiently far down this path would mean losing our ability to make any independent choices of great significance about our own destiny. What should we think about such loss-of-human-control scenarios? My goals include identifying the relevant notion of control, distinguishing genuinely valuable forms of self-determination from mere anxieties about the loss of importance, and sketching a vision for a future in which humanity retains control in ways that matter.