What racing reveals about working with AI — the OpenAI Podcast Ep. 22

TL;DR: Motorsports reveal how AI (like ChatGPT, Codex) can boost team performance through data organization, real-time analysis, and educational collaboration — with the focus on a human-AI "soft interface." ## Background This episode of the OpenAI Podcast, hosted by Andrew Mayne, features two guests: OpenAI researcher Joyce Rapiel and Chase Holden, co-founder of RaceTek Systems. Joyce discusses a research collaboration with Chip Ganassi Racing in the IndyCar Series; Chase tells the story of using ChatGPT and Codex to build a racing intelligence software company. ## The Data War in Motorsports "Data is everything in motorsports today. I believe we're in what's called a 'data war.'" Joyce notes that racing generates massive amounts of high-frequency time-series data, requiring both retrospective analysis and real-time processing for quick decisions. Chip Ganassi Racing has 35 years of history, accumulating vast datasets from different drivers, tracks, series, and cars. The key is making that knowledge easy to access, compare, and validate. "The faster you can extract the answers you need from the data, the more experiments you can run, and the finer the tuning you can do before each race." Chase's company, RaceTek Systems, focuses on building racing intelligence systems that centralize various information onto one platform, helping teams make better decisions every race weekend. "It always comes down to who can organize the best." He emphasizes that whoever can integrate information fastest and best gains an edge throughout the season. ## From Fan to Founder: Chase's Journey Growing up in South Louisiana, Chase fell in love with racing at age five after a family friend took him to Talladega Superspeedway. He later went into media, but by 2017 felt the need to do something more meaningful, so he started a podcast. When ChatGPT launched in 2022, he realized he could finally turn his racing ideas into reality: "I had a way to build it." Before, people dismissed his ideas as "impractical," but AI tools let him do it on his own. He wanted to bring something new to motorsports, especially helping under-resourced mid- and back-of-pack teams. ## Joyce: From Car Enthusiast to AI+Racing Joyce runs an automotive enthusiast social channel at OpenAI and fell deeply in love with racing during the pandemic, following sports car racing, IndyCar, and NASCAR. She previously worked in the automotive industry. At an industry conference held in Indianapolis, she was invited to tour Chip Ganassi Racing's shop, which sparked her thinking about how AI could help the sport. By late 2024, the team was already exploring AI applications, at a time when few were talking about it. ## AI Education: Demystifying the Tool Joyce emphasizes that one of the most impactful things they provide to the team is education: "It's education about how our models work. That helps demystify them — showing that they're essentially just machines, not magic." Once engineers understand how the machine is built, it's easier to imagine how to collaborate with it better. This manifests in prompt engineering: input determines everything, and the user is responsible for finding the right input. Most preparation with Chip Ganassi happens before race day to ensure personnel at the timing stand are fully prepared for each track. ## Breaking Down Driver Communication Barriers Chase points out that AI can help break down communication barriers on the driver side. Not every driver is an engineer; sometimes they can't fully understand the information being conveyed. With AI, teams can present what would take a driver four or five hours to study in a clear, readable, understandable format, so the driver knows exactly what to do on track — for example, preparing for the actual race based on practice reports. "Decisions become smarter. They're always making the best judgment with the information they have. So speeding up that process is the name of the game." ## From Excel to Seamless Collaboration Chase observes that the racing world largely lives in Excel. Models can work with Excel, but it's not an ideal format. Part of the work is finding ways to enable as seamless a collaboration as possible between humans and machines. "AI is the machine-side component. We need not only to make AI fit how people work today, but more importantly, to let them evolve together." People need to learn how to collaborate and communicate with models in the most seamless way possible. ## The Soft Interface: Beyond Telemetry Joyce raises an interesting concept — the "soft interface." Beyond numbers and telemetry data, there's a vast amount of "soft" information, such as a driver's description of how the car handles, or the engine's... (transcription cuts off). Traditionally, this type of information has been handled by scripts and plotting algorithms, but AI promises to better integrate this unstructured data, helping teams form a more complete picture. --- Source: What racing reveals about working with AI — the OpenAI Podcast Ep. 22 (https://www.youtube.com/watch?v=KNPjRpNtQ7s)