A paper published on the 30th in Nature reported on an artificial intelligence (AI) system called Swift, which has the ability to drive drones and defeat human opponents in real-world one-on-one championships. The research results mark a milestone in mobile robotics and machine intelligence, and can inspire deployment in other physical systems, such as autonomous ground vehicles, aircraft, and personal robots. The deep reinforcement learning system has defeated humans in various games, including chess, Atari games, StarCraft, and GT racing. However, most of these are simulations and board game environments, rather than competitions in the physical world. In the first person view of drone racing, professional athletes need to drive high-speed aircraft through three-dimensional tracks. Each pilot observes the surrounding environment from the drone's perspective through video live streaming from an onboard camera. It is difficult for autonomous drones to reach the level of professional pilots, as robots need to fly within their physical limitations and can only estimate their speed and orientation based on onboard sensors. The research team of the University of Zurich in Switzerland has designed a auto drive system that can compete in the physical world at the level of the world champion of mankind. This system combines simulated deep reinforcement learning with data collected from the physical world. In a series of competitions designed by a professional drone racing pilot on a real-world track, Swift competed against three human champions, including two world championships in international leagues. Human pilots have a week to train on real tracks, and then each pilot competes against Swift in multiple one-on-one matches. The system defeated each human champion in multiple rounds (15 out of 25 wins) and broke the record for the shortest race time, leading by half a second compared to the fastest speed of a human pilot. In the accompanying news and opinion articles, scientists commented that the system must be able to cope with external disturbances such as wind, changes in lighting conditions, unclear definitions of various obstacles, other racing drones, and so on - all of which pose a significant challenge to existing AI technology. Editor in Chief Circle: This is undoubtedly a breakthrough technology, but we should see its two sides. As an important achievement in the field of AI, this system can be further developed in a more realistic and ever-changing environment, fully unleashing its application potential; On the other hand, researchers have to consider the technical and ethical aspects that AI controlled drones should possess - including but not limited to software vulnerabilities, permission errors, and potential challenges for human decision-making power. Whether as military weapons or civilian flights, we must take precautions against abnormal situations of drones driven by AI. (New News Agency)