Artificial intelligence (AI) is at an inflection point and is poised to move into the open world and into our lives in numerous ways that will have numerous influences on people and society. While AI promises to provide great value, along with the promises and aspirations come concerns about potential costs, failures, and rough edges. Concerns include failures of automation in safety-critical domains, influences on jobs and economy, use of biased data and algorithms, and runaway AI. Eric Horvitz addresses short- and longer-term challenges and touches on reflections and studies including the AAAI Asilomar Study and the One Hundred Year Study on AI at Stanford University.