Drone-racing human pilot defeats AI -- at least this year

A human defeated a computer-piloted drone in the Drone Racing League's first man-vs.-machine competition. But the programmers behind the AI-powered drone, Team MAVLab from the Netherlands, can take consolation in a $1 million prize for creating the fastest self-piloting aircraft this season.

The Drone Racing League mostly involves human pilots, but this year it began its first computer-piloted drone competition, called Artificial Intelligence Robotic Racing (AIRR). In it, quadcopters with dual Nvidia Xavier processors and four cameras, must navigate a curving course around obstacles and through gates. Team MAVLab won that competition Friday at the final competition in Austin, Texas.

Next came the faceoff against one of Drone Racing League's top pilots, Gabriel "Gab707" Kocher. He won handily with a 6-second trip through the course, nearly half the time of the AI-powered drone that finished in 11 seconds.

The league, though, expects the AI drones to improve and defeat humans in coming years. "2023 is our bet," said Ryan Gury, DRL's chief technology officer, in an interview just before this year's season began.

The competition spotlights the growing capabilities of computer-piloted machines. Tesla Model 3 cars are far from perfect under AI control, and self-driving cars from companies like Cruise and Alphabet's Waymo are only operating in limited tests, but the technology continues to advance rapidly. Drone maker Skydio sells quadcopters that follow a human pilot's navigation instructions, but are smart enough to avoid obstacles like trees and buildings.

Aerospace giant Lockheed Martin underwrote the $1 million prize.