The Paris Olympics will feature several AI-influenced elements, including an app to guide athletes.Credits: Chesnot/Getty
As more than 10,000 athletes from around 200 countries gather in Paris for the 2024 Olympics Summer Olympics This week, they’ll have a brand new, friendly – but faceless – voice to greet and guide them.
How do I get to my sporting event venue? Can I watch the opening ceremony live? Will there be a computer to referee my matches? Where can I get gifts from sponsors?
These are just some of the questions athletes will be able to ask AthleteGPT, a artificial intelligence (IA) chatbot Designed for them, it can be accessed through the Athlete365 mobile app. It will be able to “very quickly navigate thousands of pages of information and be available 24/7 to answer questions,” says Todd Harple, program manager for Olympic AI innovation at Intel Labs in Hillsboro, Oregon, which is involved in the effort.
Science and the World Cup: How big data is transforming football
The chatbot — a large language model (LLM) built using AI developed by Mistral AI, a company based in Paris and Intel Gaudi processors — are just one way AI is making its mark on this year’s Olympics, which begin July 26. Few people know about LLMs or have heard of them ChatGPT at the last Summer Games, in Tokyo in 2021. But the sprinters in Paris can only hope to match the progress made since then by AI technologies.
The International Olympic Committee (IOC) is embracing this technology. In April, it launched its AI Agenda — an effort to streamline the explosive growth of AI research in sports and develop a strategy for its use at the Olympics. “We need to be the leaders of change, not the object of change,” IOC President Thomas Bach said in Lausanne, Switzerland, at a news conference in London that showcased the capabilities of various AI sports tools.
Nature explores three ways AI is changing the way athletes and spectators will experience the Olympic Games.
Athlete Performance and Training
As early as 1900, when Paris first hosted the Olympics, French scientist Étienne-Jules Marey pioneered the use of technology to study athletes in motion. His high-speed chronophotography, which involves setting up a camera like a machine gun and feeding it photographic plates like ammunition to quickly capture images, made it possible to photograph sprinters and long jumpers. He analyzed the biomechanics of the body to “Discover the secret of the superiority of certain athletes”A Nature editorial commented in 1901.
Today, much more can be done by simply recording with a smartphone. Intel’s 3D Athlete Tracking Technology (3DAT) uses AI to track 21 points on the human body to accurately reproduce physical movements, providing “all the biomechanical information that coaches are looking for” in elite athletes, Harple says. He believes such technologies will lead to tighter competition and new records.
AI is being used in a variety of ways to improve athletes’ performance, from designing custom athletic shoes and apparel to determining optimal training and nutrition programs. “It could even accelerate our discovery of new strategies for practicing sports,” he says. The Fosbury fail, now the dominant high jump style pioneered by American athlete Dick Fosbury at the 1968 Olympics, is a historic example of such a fundamental shift.
Capturing photos and videos of movements and analyzing the data is one way athletes can improve their performance using AI.Photo credit: Hector Vivas/Getty
The ease of collecting individual data, combined with AI analysis, could also help coaches identify talent, making the sport fairer. In March, the IOC piloted a screening program that used 3DAT to identify more than 40 children in Senegal who showed promise to become Olympic athletes, by analyzing simple exercises like running and jumping.
But sports and nations with major professional leagues will maintain a significant advantage because they have the resources to collect high-quality data and train algorithms based on it. “The problem with some Olympic sports is that there’s not a lot of data,” says Patrick Lucey, chief scientist at sports technology company Stats Perform in Chicago, Illinois. That expands how the technology can be used in other aspects of these games, such as judging and refereeing.
Arbitration and real-time data
Frank Ohme, an Olympic water polo referee, is no stranger to artificial intelligence. His day job as an astrophysicist at the Max Planck Institute for Gravitational Physics in Hanover, Germany, involves searching for signals of black hole collisions — sometimes with the help of artificial intelligence — in noisy gravitational wave data. But when he dons his all-white referee uniform in Paris, he’ll have to peer through splashes of water to determine whether the ball crossed the finish line into the goal.
AI already informs such decisions in sport, including footballusing information recorded by a set of cameras installed in the stadium and chips implanted in the ball. But this technology has not yet made its way into other sports, and AI will probably be slower to take hold in areas such as refereeing, which requires real-time data analysis.
Another hurdle is funding and the specific needs of each sport – there will be 32 events at the Paris Games. Although water polo is the oldest Olympic team sport, there isn’t as much money to spend on it as there is for basketball or soccer, Ohme says. Using AI in water polo would also present different challenges, such as training the algorithms on underwater footage and in chaotic scenarios, he adds.
Accurate and open communication is essential when using AI to make real-time decisions. “The easiest way to convince teams and viewers is to give them all the information through an image or visualization where they can determine (the decision) themselves,” Ohme says.
What ChatGPT and Generative AI Mean for Science
It’s also difficult to disambiguate actions like fouls in contact sports. These are split-second decisions that even humans can’t agree on. “I wouldn’t even know how to quantify them,” says Ohme, who believes detecting black holes is an easier task for AI by comparison.
Improving the viewer experience
The data streams collected during the Games will feed not only AI algorithms, but also viewers hungry for statistics. “Sport is a language in itself. It crosses barriers to help everyone communicate,” Lucey says. Statistics and numbers enrich these conversations by providing additional benchmarks for comparison. “Of course people want that,” he adds.
Broadcasters are racing to find ways to augment this new wealth of information and get it onto TV screens. Viewers were mesmerized when the virtual world record line was superimposed on the screen for viewers during the Sydney 2000 Games. By 2024, broadcasters will have the ability to display much more information, such as acceleration, top speed and stride length, Harple says.
What excites Harple most is the prospect of providing viewers with personalized highlights using Intel’s Geti computer vision platform, which could be a feature of future broadcasts. With so much sporting action being recorded at once, Harple says AI’s ability to select exactly what viewers want to see will be a game changer. This could be particularly beneficial for coaches and broadcasters in countries that have more limited access to production resources. “If someone wants to see every three-point shot by the Nigerian men’s basketball team, AI can go through all the footage and automatically stitch it together,” he says.
