Jay Hirano/Shutterstock
We live in a time when AI is moving into our world like wildfire, quickly closing the gap between human ability and machine precision. Watching professionals face off against "infant AI" has become surprisingly thrilling. And now, that showdown has reached the world of single-seat F1-style racing through A2RL, the Abu Dhabi Autonomous Racing League at the famous Yas Marina circuit.
Testing the limits of autonomous racing, eleven teams competed for the title with their cars piloted entirely by advanced AI. In the end, the German team TUM clinched the championship, securing their second consecutive victory. In an unprecedented feat, TUM's AI even outpaced the former Russian F1 driver Daniil Kvyat in the Man vs AI showdown by 1.58 seconds. Even Kvyat acknowledged how far things have come, noting that the gap compared to just a couple of years ago is massive. He admitted that the progress has been genuinely impressive and that being on track with an AI driver felt unique and even exciting, as the two delivered a proper battle for the fans.
Moreover, compared to the previous year, which was also the opening season of A2RL, this time saw far fewer crashes, fewer unexplained halts, fewer spin-outs, and overall a much smoother showing from the autonomous cars. Important for the organisers as high-class Emiratis attended the event. Interestingly, Audi had attempted a similar "man vs machine" idea a few years ago with a self-driving RS7 racing against professional drivers. However, back then, the autonomous car was still noticeably slower than an experienced racer, showing just how quickly the tech has evolved.
Future of racing?
Peepo/Getty Images
AI in racing may be a very new concept, but if we think deeper, it has been here in racing video games for quite a while now. Many racing games include advanced AI models that are quick enough to beat even the best gamers at their hardest settings. "Forza Motorsport" in particular is known for having really good AI. But there is a huge difference between AI in video games vs AI in real life.
Similarly, advancements in computational power as well as groundbreaking new AI tools have now led to AI being bright and fast enough to not only make the right decisions but also do so as quickly as peak humans. The average speed at the event was 250 kmph (155 mph). For reference, Waymo's Level 4 autonomous taxi has a top speed of 65 mph on freeways, three times the speed requires more than three times the computational power.
However, it's not without its issues. During any racing event, crashes are expected, and there were a few crashes at this one too. Fortunately, the F1 driver's years of racing experience helped him avoid any incidents during his run against TUM. But when competing with actual drivers, if something were to go south, one couldn't hold AI accountable for its actions. So we believe autonomous series in the future would not include any human drivers. Instead, autonomous leagues of the future could be the playground for groundbreaking science projects and advanced AI technologies, think Microsoft and OpenAI rather than Ferrari or Mercedes.