By hijacking a community tool based on machine learning, cheaters can beat excellent players without even touching their controller.
It’s been a long time since AI-based systems have been able to humiliate humans in lots of games. Whether it’s chess, go, Texas Hold’em, or DOTA 2, a great classic of e-sport, these algorithms no longer give flesh and blood beings a chance. And it’s starting to become a problem for the publisher of the hugely popular Rocket League.
A few years ago, these confrontations between humans and machines were above all scientific experiments intended to test the performance of these algorithms. They were all organized by companies with significant financial resources that employ great specialists with rare skills.
But over time, the general public has also become familiar with these technologies. Today, even if they remain the undisputed leaders, machine learning is no longer reserved for the titans of the discipline. Individuals have started developing their own AI-powered bots — and unlike DeepMind and the like, it’s rarely basic research.
Some ill-intentioned developers have started programming bots with one radical objective: to cheat in competitive games to reach the top of the leaderboards. And in recent months, it’s been the elite of Rocket League who paid for it.
Nexto, a diabolical dribbling bot
This problem, already well known within the community, exploded into the open after the publication of an article by Wired. The American media reported the words of Reed “Chicago” Wilen, who is currently wearing the G2 Esports jersey. With several first places in major events and more than $450,000 in cumulative earnings, the 21-year-old American is one of the stars of this competitive scene.
The interested party is therefore used to facing very high level opponents. But in one area in particular, namely the dribblenone of them come close to Nexto, a bot with superhuman abilities.
Rocket League is a physics-based game that is very easy to get into, but immensely difficult to master from a technical standpoint. To reach the highest levels, players must learn to maneuver their vehicle to delicately move a huge ball while multiplying acrobatics, while remaining aware of their surroundings at all times.
Some players have specialized in this aspect of the game. After thousands of hours of training, they have acquired tremendous control over the ball. But this remains disproportionate to the diabolical efficiency of Nexto. “ His perfect dribbling would wreak havoc on any player “, explains Reed Wilen to Wired.
A training program hijacked by cheaters
Originally, Nexto was designed as an open-source resource by RLGym, a well-meaning community of fans. The objective was to make it a tool at the service of the community. For example, he can serve as a sparring partner to help advanced players work on their ability to perform or defend against elaborate dribbling. He had absolutely no vocation to play real matches.
But apparently, smart guys managed to cheat by hijacking Nexto to play for them in competitive games. Result: players have approached the highest ranks of the ladder without even touching their controller. A heartbreaker for those who have spent hours mastering the basics of the game.
Luckily, even though Nexto’s dribbling is terrifying, he still falls far short of the pros in other game compartments. So if Chicago is to be believed, he doesn’t threaten the supremacy of absolute elite players… instant.
Because behind the scenes, among the developers of the legitimate version of Nexto, the bot continues to gain momentum. Before long, it could become just as good in the other compartments of the game. If hackers manage to hijack these new versions, it could spell disaster for the integrity of the competition.
” Nexto is already superhuman in some situations “says Zealan, a Nexto developer interviewed by Wired. “ Trust me, in a few years he will be way above the pros. “. ” When it does, it won’t be a cakewalk for other Rocket League players. “says Chicago.
To cut it short, Psyonix therefore took the lead. The game’s publisher has launched a major cheat-hunting campaign. According to polygon, developers are preparing a host of tools to track them down. With particular attention to those who use super-bots. For example, they could train their own neural networks to recognize these intruders. Hope for the players of Rocket League that they will succeed, and especially that the other publishers of competitive games will know anticipate the rise of cheating boosted by machine learning.