AI against a professional pilot

The profession of the “quadrocopter pilot” appeared quite recently, and the artificial intellect seems to be already trying to take it away from the person. However, until it failed.

The NASA Jet Propulsion Laboratory tested autonomous unmanned aerial vehicles for two years and arranged a race between the man and the AI ​​as the reporting event for his experiments. The good news is that the race was won by a man – but only after hard training.

The research was funded by Google, and test drones, called Batman, Joker and Nightwing, used Tango technology to analyze the environment in 3D. In the tests took part drones, which could fly at speeds up to 130 kilometers per hour.

(Although during the race the speed of the vehicles did not exceed 80 kilometers per hour).
Against the AI ​​NASA came the professional pilot Ken Loo, who flies under the nickname FlyingBear and participates in the international league Drone Racing. According to NASA, AI initially defeated Loo, but only until the pilot thoroughly studied all the features of the route.

“It’s definitely the most difficult track, of all that I’ve ever flown,” Loo said. “One of my drawbacks as a pilot is that I easily get tired.” And when I’m tired, I’m starting to get lost, even if I’ve already flown by this track 10 times. ”

Of course, AI did not get tired, but he lacked intuition. “Our algorithms resisted a man who flies at the behest of his heart, acting intuitively,” NASA press statement said. – AI during the passage of the route chooses the optimal angles and speeds, while the man acts more aggressively, unexpectedly, risky. ” And man wins. Till.

What conclusions can be drawn? Autonomous drones are already used for a number of tasks requiring routine and careful flight, for example for surveillance and delivery. Probably, AI will soon be able to “drive” as quickly as a person. However, one should take into account such complexities as, for example, blurring of the image during movement, which limits the recognition abilities of cameras and sensors of the drone. However, this is a matter of time and technology.

Notify of
Inline Feedbacks
View all comments
Would love your thoughts, please comment.x