Progress Toward Athletic Intelligence [From the Editor’s Desk]

Abstract

We have witnessed extraordinary advancements in generative artificial intelligence (AI) in 2023. Whether it’s composing essays or generating pictures based on written descriptions, it appears that machines could eventually achieve general intelligence matching, even exceeding humans on some tasks. But those tasks do not include robotic ones yet, which require physical skills and the integration of perception and actuation. A breakthrough was made in robotics in that an autonomous system won against human world champions in first-person-view drone racing [1] . Using an innovative hybrid learning-based method, the robot’s onboard perception system translated high-dimensional visual and inertial information to low-dimensional representation, and a control policy was trained that integrated perception and control commands. The autonomous system achieved the fastest race time against three drone-racing world champions. The progress is encouraging, and we hope to see more autonomous robots that can reach human-level abilities and even compete against humans in sports (such as soccer or tennis) and other desirable tasks (such as cooking or folding clothes).