Some day, there may be an Artificial Intelligence Olympics, where generative AIs face off to create the wildest images, videos, and hallucinative text responses. In the meantime, we have the real Olympics, starting July 26 in Paris, France. They will feature flesh and blood athletes, but behind the scenes will be AI chatting with athletes, chunking up highlight reels, and telling attendees which Olympic sport they’re best equipped to compete in (at least in their fantasies).
“Our big focus for these games, which makes sense as it’s really the hot topic right now, is artificial intelligence, and so we are the official artificial intelligence platform provider for the Olympic Games,” said Sarah Vickers, who leads Intel‘s Olympic and Paralympic Games Office.
I spoke to Vickers just as she was about to fly off for Paris, where she’ll spend the next two months overseeing more than 100 Intel employees building and integrating Intel’s Olympics AI strategy.
Unlike previous years when Intel once invested in drone intelligence to help build some of the earliest Olympic drone synchronized flying shows, Intel’s 2024 Paris Olympics work will remain grounded and focus on a handful of key areas.
Medaling in chat
Yes, Intel is using its Gaudi 2 generative AI platform to power an Olympic chatbot that will help more than 10,000 athletes navigate the sometimes confusing Olympic Village and games.
Vickers told me it’s an LLM for athletes and explained it will be integrated into the Athlete 365 application they already use. It should help stressed-out athletes figure out the Olympics’ day-to-day operations and what’s expected of them. “Just understanding all that stuff that they have to search or find in PDFs; putting that in a conversational chatbot for them that’s integrated into an app that they already have and that’s running online as well.”
It might quickly answer if taking a Tylenol before an event could present a doping problem or simply the bus schedules for getting to and from the various sports venues.
With athletes from 200 countries, I was curious if the chatbot could speak multiple languages. Vickers wasn’t certain yet if it could. There are also no explicit plans to open the chatbot up top the world, though Vickers thinks Intel may look for a way to showcase it to consumers not attending the games.
As for what safeguards the LLM includes, Intel’s Responsible AI group contributed to the chatbot’s development. “I don’t think you can stop people from asking crazy questions, but I think you can stop what it responds to and use that to identify that this is something we don’t have the answer to, and then maybe figure that out,” Vickers told me.
The bronze in broadcast
On the viewer (and producer side), Intel AI will be combing through event footage to cull and present highlights. If you think there are already enough video highlights online, just imagine little-watched events that, while being taped, may not get much broadcast or production love. Good luck finding highlights of an athlete nailing his tenth perfect target in Clay Pigeon Shooting or an emotional shot of his parents rooting him on.
“So highlights, traditionally, when Olympics Broadcast Services has done them, they were limited by the number of people they had to create highlights. They didn’t have a lot of automation. They did some tagging, but all that tagging was done manually, and so it was just there was only so much they could put out. Now with AI that we’re doing, auto-tagging, the ability to create real-time highlights, which can be done in just a few minutes and almost real-time.”
This will not only help obscure sports but also smaller countries that can’t afford to send full broadcast teams to find and present region-based highlight reels. This AI tagging might make it as simple as choosing the sport(s) and the country, and then hitting an export button.
Gold medal skills
Image 1 of 3
If you’re lucky enough to attend the Olympics, you’ll see AI at work in the Stade De France, which is France’s largest sports stadium and will become the Olympics Stadium for the games. Intel and Samsung have partnered to offer an AI-powered talent identification system on the ground. Using Samsung smartphones and tablets and computer vision to capture participants doing a handful of sports-related exercises, Intel’s cloud-based AI will identify the Olympic sport they’re best equipped to handle.
The two tech companies piloted this program in Senegal, where they used it to help identify local youth as possible athletes for the Youth Olympic Games, which will be held in Senegal in 2026.
“We went and had 1000 children participate in this activity to see…what is their skill? What are they able to do? And that was all able to be done on with a mobile phone, [and] very low infrastructure costs on the ground,” explained Vickers.
Silver in navigation
Other AI integrations include a wayfinding system originally designed for the vision impaired, which will now assist everyone in finding their way through and around the Olympic Village. AI is also used to analyze space use across the media lounge and village.
“If you think about just people movement, if you understand occupancy, you understand better how to manage food and beverage. We understand better how to manage the queues for transportation. Using that data to help just make better decisions. That information is also being used by the IOC [International Olympic Committee] to help plan for the future.”
Most of Intel’s AI integrations run in the cloud, usually with Intel Xeon processors. The software is usually Intel’s Geti applications, which helps with performance and model building. Vickers told me some of the AI could run locally (or on the edge) in devices like Samsung tablets and phones, but that’s not how it’s being done for these Olympics.
What Intel is not doing is bringing any of this AI directly to the opening ceremonies. For that pageantry, it’ll just be humanity, emotion, and whatever stagecraft the producers can apply – and probably drones, but not supplied by Intel.
You might also like
https://cdn.mos.cms.futurecdn.net/3Z2v9PPs4WZXcYG35kH7C4-1200-80.jpg
Source link
lance.ulanoff@futurenet.com (Lance Ulanoff)
