Back in 2016, AI shocked the world when AlphaGo beat a human champion, and in this video I’m talking about what feels like the next big turning point—AI proving itself in the physical world. At the A2RL autonomous drone championship in Abu Dhabi, drones running onboard AI systems went head-to-head against elite human FPV racing pilots, and the AI actually came out on top in multiple events. The standout was a MavLab drone from Delft University of Technology that used a deep neural network with high-frequency onboard processing, combining vision and motion data without motion capture or external tracking. What’s wild is this wasn’t some university sim on a computer—this AI flew a real drone through a complex indoor course at speeds pushing around 100 km/h, ripping tight turns and narrow gates, and it beat three world champion pilots in a knockout round. But it’s not a total takeover yet: in a later best-of-nine, a human pilot edged out an AI competitor, which tells me we’re in a transitional era where humans and machines are pushing each other. For us FPV pilots, this points toward smaller onboard autonomy modules and what that could mean for future features—navigation, obstacle avoidance, long missions, and maybe even something that eventually shows up in Betaflight as a preset. Personally, I still love solid manual acro with no AI “help,” because that’s what makes FPV fun.