Neural network beat humans in drone race

No time to read?
Get a summary

The Swift artificial intelligence system, developed by researchers from the University of Zurich and Intel representatives, broke the record for the fastest completion of the course by defeating the world champions in drone races. About this “Moscow 24” said scientific observer Nikolai Grinko.

For this experiment, a 25 x 25 meter rail was built in an aircraft hangar near Zurich, on which seven doors and one vertical turn were placed. The neural network, which was trained in high-speed drone piloting before the competition, was challenged by two world champions in the sport, as well as a three-time Swiss champion in street racing. According to them, the artificial intelligence system performed “completely inhuman maneuvers”. Swift’s fastest lap was half a second faster than her best human score.

But the experiment showed that humans are better able to adapt to changing conditions. So when the bright sunlight illuminated the hangar more than planned, the neural network failed.

At the end of July, the Ministry of Sports of Russia accepted and he included drone racing and laser warfare in his All-Russian Sports Records.

Earlier it was known that there were students of Moscow University. allowed Use neural networks as you work.

No time to read?
Get a summary
Previous Article

The expert announced the removal of Reznikov from the post of head of the Ministry of Defense of Ukraine

Next Article

Calls on Russians to install anti-spam protection on smartphones