Abstract:
In this essay, I propose three novel challenges that I hope will encourage progress toward human level cognitive systems. Each problem – one in entertainment, one in law, and one in politics – involves the integration of components for which initial technologies exist, although the component tasks remain difficult in their own right. Each challenge also revolves around a virtual embodied agent that interacts with human avatars in simulated environments. In each case, I describe the overall challenge problem, subtasks on which researchers can make independent progress, graded versions of the problem that would enable incremental improvement, and methods for evaluating the resulting cognitive systems. I also consider the reasons that both researchers and the public will find the challenge tasks interesting and worthwhile.