The Robot Judges of Gymnastics
The camera pans across the crowded arena as Simone Biles, the undisputed queen of gymnastics, takes her place on the vault runway. The air crackles with anticipation as she explodes off the springboard, a breathtaking blur of muscle and grace, twisting in the air before landing with a perfect stick. But there are cameras watching Biles that you, the spectator, likely don’t see: four strategically placed high-definition cameras, capturing every angle of her vault, her body rendered as a precise 3D model. These cameras are part of the Judging Support System (JSS), developed by Fujitsu and FIG, the International Gymnastics Federation, to help judges make more accurate and consistent rulings.
While Biles’ performance was a display of unmatched skill, the JSS was originally designed to resolve those moments where judging gets a bit blurry, where tenths of a point can separate a gold medal from a disappointing fourth place. For years, the dream of AI-assisted judging has simmered, fueled by the belief that humans, with all their inherent flaws, cannot perfectly judge the intricate ballet of gymnastics. The quest to create a system that could not only identify but also evaluate movement with the precision of a machine has been a long and winding road for Fujitsu and FIG.
It started as a joke, explained Morinari Watanabe, the first Japanese president of the FIG, in a press conference at the 2023 World Gymnastics Championships in Antwerp. Watanabe, in 2015, had jokingly suggested to Fujitsu that they develop robotic judges for gymnastics, but his jest sparked a serious initiative. The project took six years of development and tens of millions of dollars before the JSS was finally ready for prime time in Antwerp.
However, the JSS is not intended to be a replacement for judges; for now, it’s a second opinion, a tool to be consulted in cases of inquiry. This begs the question: can AI truly judge human excellence better than a human?
The JSS captures gymnast movements with incredible precision, producing 3D renders that deconstruct every twist, turn, and tuck into tangible data points. Angles, distances, and timings are meticulously recorded, offering an objective measure of execution and difficulty. But while the JSS excels at analyzing individual elements, it struggles with the nuances of judging a complete routine, like the intricate dance between connection value (CV) and the smooth flow of movement on the beam. This subtle element, where gymnasts link skills for bonus points, currently escapes the JSS’s numerical lens. The JSS, with its mechanical precision, cannot yet grasp the subtleties of human expression in movement.
Despite this limitation, the JSS has already proven its efficacy in resolving edge cases, instances where the human eye struggles, like determining if a gymnast has reached a perfect handstand during a dismount. This was illustrated during the 2012 London Olympics when Kohei Uchimura, a legendary gymnast, narrowly missed a handstand on the pommel horse, jeopardizing Team Japan’s medal chances. The judges, lacking the precise measurements the JSS offered, were forced to make a subjective call based on what they could perceive with the naked eye. The JSS could have potentially provided the definitive answer, ensuring a more accurate and objective decision.
Even with potential benefits to judging, the JSS is a costly endeavor. The system requires extensive hardware and software, along with a dedicated team for setup and monitoring. This raises concerns about its practicality for smaller, less resource-rich competitions. However, Fujitsu is working on a more accessible version for training purposes, believing that the JSS’s ability to analyze minute movements could be used to prevent injuries and enhance performance.
Beyond its applications in gymnastics, Fujitsu is aiming to leverage the Human Motion Analytics (HMA) technology developed for the JSS across various industries. The ability to accurately capture and analyze human movement holds potential for sectors like healthcare, ergonomics, and security, transforming the system into a multi-faceted technological tool.
This raises deeper questions about the future of sports judging and the ethical implications of AI. While the JSS seeks to improve accuracy and consistency, its use also brings to light concerns about privacy and data security. The system collects highly detailed information about athletes’ movements, raising questions about how this data is used, stored, and protected.
The pursuit of objectivity in sports often leads to a push for technological solutions, but the JSS’s development highlights the delicate balancing act between human artistry and cold, hard data. While AI can be a powerful tool for analysis and decision-making, it lacks the ability to truly appreciate the artistry and human spirit that drive competition. The JSS, in its quest to perfect the judging process, may unintentionally create a world where gymnastics becomes another data point for a machine rather than an expression of human potential.
Ultimately, the JSS’s success hinges on embracing its limitations while highlighting its strengths. As AI technology advances, it can offer valuable support to judges, providing objective reinforcement for their expert opinions and resolving complex judgment calls. However, the human element of judging will always remain crucial, as it is the human capacity for appreciation and understanding that allows us to truly savor the beauty and athleticism of gymnastics.
