American gymnast Simone Biles at the 2016 Summer Olympic Games in Rio de Janeiro. Image via Ben Stansall/AFP/Getty Images

Gymnastics Officials and Tech Company Hope to See AI-Assisted Scoring At 2020 Tokyo Games

According to an article published in The Guardian, the International Gymnastics Federation (FIG) is considering using AI technologies to help judges analyze gymnasts’ performances in the upcoming 2020 Olympic Games in Tokyo. By analyzing data from an array of 3D sensors, FIG hopes that the device—slated to be funded and built by the Japan-based tech corporation Fujitsu—will introduce an unprecedented degree of objectivity and precision in scoring, helping to reduce the influence of bias, corruption, and other human factors on athletes’ scores and providing the safeguard of a trusted independent method of reviewing a gymnast’s performance should the judges come into conflict.

A promo video demonstrating Fujitsu’s gymnastics technology in action. Via YouTube/Fujitsu.

Leading the project is FIG’s technology chief Steve Butcher, who reassured critics of the proposal by stating that “this is not to replace judges.” Instead, he says the technology is a welcome complement to the hard work of Olympic judges: “in judges, it’s proven there can sometimes be an inherent bias. A program like this can be helpful for settling issues or technical questions.” Prior FIG president Bruno Grande shares Butcher’s optimism about AI’s ability to provide fair and objective results in contrast to relying solely on judges who “must work for eight hours per day— does that allow the mental capacity to remain coherent? It’s not possible to maintain a coherent mind of criteria. Only the computer does.”

The program is not without concerns and detractors, however. In conversation with the Guardian, Olympian Nadia Comaneci commented that “I don’t think it will be possible to ever replace judges. Gymnastics is too complex, there are so many skills and nuances in every routine,” sharing concerns that Fujitsu’s approach might detract from the innovation and artistry of cutting-edge gymnastics or even be unable to detect and accurately score novel approaches and maneuvers if  it were trained instead on more commonplace gymnastic routines: “gymnasts are known for pushing the skills, looking for new angles, turns, points—so what happens when someone comes along with a totally different routine that has not been seen or registered by the computer, how would that be judged?”

Even more worrisome though is the threat of tampering and hacking. While Fujitsu and FIG tout the AI’s ostensible objectivity and physical supremacy to human judges as reasons to adopt the technology, critics fear that the device is far too vulnerable to exploitation and that nations or individuals could find imperceptible ways of hacking any officially-sanctioned AI scorekeeper.

UC Berkeley’s executive director for long-term cybersecurity Betsy Cooper, author of The Cybersecurity of Olympic Sports: New Opportunities, New Risks defined some of the potential pitfalls as follows: “In gymnastics, you can have 10 to 100 independent moves the system is trying to score. If the algorithm were manipulated by even a small portion you could affect the overall outcome score and it would be very hard to detect,” saying that one could “manipulate the algorithms to change the score one out of every five times, making it hard to detect […] whoever has an interest in the outcome of these major sporting events will also have an interest in trying to take advantage of any such system.”

While the AI has yet to be officially approved for the Tokyo Olympics, the device will see its debut at the gymnastics world championship in Doha, Qatar, and if it does well then Fujitsu and FIG will likely campaign for it to be included among the highly anticipated technological wonders slated for exhibition at the Tokyo games, described by the Guardian as “the most technologically advanced Games in history.”