In a fascinating twist of human versus machine, artificial intelligence (AI) has once again demonstrated its prowess, this time in the exhilarating world of drone racing. Picture this: three skilled drone racers, experts in their craft, found themselves pitted against a formidable opponent – an algorithm known as Swift, developed by researchers at the University of Zurich. This AI marvel took on the challenge of navigating a 3D racecourse at mind-boggling speeds, all while aiming to avoid crashing. And it did so with surprising success, winning an impressive 15 out of 25 races against these world-champion human pilots.
The drone racing arena is no place for the faint-hearted. Drones zip through the course at breathtaking speeds of up to 50 miles per hour (around 80 kilometers per hour), subjecting both the machines and their human counterparts to accelerations reaching up to 5g – an intensity that could cause even the bravest souls to black out.
What makes this achievement even more remarkable is that Swift, the AI-powered drone, managed to outshine human champions in a sport designed exclusively for humans, marking a groundbreaking moment in the world of robotics and sports.
So, what exactly is drone racing? In this thrilling competition, pilots steer their drones through a racecourse filled with gates that they must navigate with precision to avoid collisions. Pilots get a first-person view of the action, thanks to a camera mounted on the drone, which transmits live video feeds to their goggles.
Do you know that the Gannett newspaper chain has stopped using an artificial intelligence tool to compose high school sports articles? This decision was made after the AI technology made some embarrassing errors in articles, causing a commotion in at least one Gannett-owned publication:
The competition was no mere fluke. Prior to the showdown, the human pilots were given a week to practice on the course, honing their skills and strategies. In stark contrast, Swift underwent rigorous training in a simulated environment, complete with a virtual replica of the actual racecourse.
The secret to Swift’s success lies in a technique known as deep reinforcement learning. This approach involves trial and error, meaning that during training, the drone crashed hundreds of times. However, because this was all happening in a simulation, researchers could simply hit the reset button and start anew.
During an actual race, Swift relies on its onboard camera to capture video, which is then processed by a neural network. This network detects the racing gates. Concurrently, readings from an inertial sensor help estimate the drone’s position, orientation, and speed. All these data are then fed into a second neural network, which calculates the precise commands to steer the drone through the course.
Analyzing the races, researchers found that Swift consistently excelled at the race’s outset, executing tighter turns than its human counterparts. Its fastest lap clocked in at a blistering 17.47 seconds, a whole half-second quicker than the speediest human pilot. But, it’s important to note that Swift was not infallible. It did lose 40% of its races against humans and experienced occasional crashes, particularly in response to environmental changes like shifts in lighting.
Do you know how the secrets of GTA 6 were stolen? Imagine you are a diehard fan of the GTA series, avidly awaiting any information regarding the upcoming GTA 6. Until a young hacker named Arion Kurtaj appeared on the scene and unleashed a torrent of leaked GTA 6 videos online, the wait seemed intolerable:
The outcome of these races left the world champion pilots with mixed emotions. As Thomas Bitmatta put it, “This is the start of something that could change the whole world. On the flip side, I’m a racer; I don’t want anything to be faster than me.” Marvin Schäpper added, “It feels different racing against a machine, because you know that the machine doesn’t get tired.”
What makes Swift’s achievement even more remarkable is its ability to handle real-world challenges, such as aerodynamic turbulence, camera blur, and variations in lighting. These are obstacles that can often bewilder systems reliant on pre-computed trajectories. This suggests that Swift’s capabilities could extend to tasks like search and rescue missions in burning buildings or inspections of massive structures such as ships.
The military has a keen interest in AI-powered drones, but experts are cautious about assuming that these advancements can be seamlessly integrated into military contexts, particularly in drones and autonomous weapon systems used for critical tasks like target selection. While AI undoubtedly holds potential for military applications, its exact role remains a topic of debate.
In the end, Swift’s triumph against human drone racing champions is not just a victory for AI but also a glimpse into the future of human-machine interactions, where AI complements and challenges human abilities in remarkable ways.