4 March 2017

(Automated) planning for tomorrow: Will artificial intelligence get smarter?

Edward Moore Geist

Edward Moore Geist is a MacArthur Nuclear Security Fellow at Stanford University's Center for International Security and Cooperation (CISAC). Previously a Stanton Nuclear Security Fellow at the...


Artificial-intelligence (AI) researchers have made very considerable advances in their theoretical knowledge of planning over the past few decades. But the impact of AI on society in the coming years will depend on how much these discoveries improve the real-world performance of automated planning, or AP, an AI subfield that seeks to create computer programs that can generate plans to achieve a particular goal. If practical applications of automated planning continue to stagnate, it could hold back all of AI, even as its other subfields continue to mature. Modest progress, meanwhile, would facilitate modest economic and military uses of artificial intelligence. And should AP experience the same kind of spectacular breakout as reinforcement learning, which is being used practically in a wide variety of fields, from robotics to finance, the peril and promise of artificial intelligence might be fully realized.