ChatGPT’s viral success since its launch has propelled interest in artificial intelligence to new heights. It has also prompted many industries to evaluate the impact that the advancement of AI technology will have on them over the next few years – both in terms of how it can be used successfully and where it will replace humans could.
Formula 1 is not spared either. The teams are already using AI in many areas, such as vehicle configuration, development direction and resource allocation. It also plays a role in planning race strategy, and this has sparked debate as to whether or not the robots could take over the pit wall entirely.
Because after a season in 2022, which has once again proven how crucial strategy is for victory in Formula 1 and how high the penalty can be if the pit wall makes a mistake, it is clear where the appeal for a successful AI model could lie.
After all, it should not collapse under pressure. It should, in theory, provide the correct answers based on the analysis of a dataset far larger than any human could ever manage, and it wouldn’t have to worry about what the press wrote about it the next day.
But what sounds so simple in theory is much more difficult to implement in practice. McLaren has well understood the opportunities and challenges with recent developments to improve its own strategy and software.
Interestingly, it’s the work the Woking-based team has done with their technical partner Splunk for their esports team that has allowed them to experiment a little more with new strategy ideas – and the potential use of AI.
Since working with McLaren in early 2020, Splunk has supported the team by providing software capable of searching, monitoring and analyzing machine-generated data in various areas of racing.
One area where Splunk has particularly excelled is in its simulation systems and, more importantly, in providing live tools – such as the famous race tracking – which are now essential for understanding the rhythm of a Grand Prix and the best pit stops for victory to plan.
Last year, as the McLaren Shadow Esports team ramped up their efforts to win the Formula 1 Esports Series, Splunk adapted a version of their Formula 1 strategy tools to help.
Being able to access the kind of data and software the team receives at the pit wall during Formula 1 races – such as analyzing tire wear and predicting under- and overcuts – proved crucial to the ambition. Lucas Blakeley won the championship for the first time and wasn’t ashamed to say that Splunk strategy software was so important to his success.
“This extra layer of depth that we’ve had thanks to Splunk, and I’ll keep saying it, is just one of the coolest things to leverage,” he says. “It just gave us that extra perspective and that extra layer. An extra arrow on the bow if you want.”
From McLaren’s perspective, the value was clear. As Ed Green, Head of Commercial Technology at McLaren Racing, explains: “It was an absolute game changer. I think it was pretty crucial to the number of wins we got that took us to the championship.”
As the Splunk system became a core part of the Esports team, it opened up avenues for some experimentation to try new things that wouldn’t have been possible in the setting of a normal Grand Prix weekend. James Hodge, Splunk’s Chief Strategy Advisor, says, “There’s more that esports can do in terms of rapid development.”
“The missions are very different. In real Formula 1, when you touch something, you’re touching an almost mission-critical system. If a Formula 1 car doesn’t send telemetry data back to the McLaren garage, they won’t be able to start the engine.”
“On the esports side, the impact is less: if you don’t get the telemetry and it’s not working, you know Lucas can still drive. Also, it’s less complex. You don’t have to put one IT facility in 20 different places around the world.”
“This way you can quickly develop prototypes. We’ve been able to try things on the game side that might have taken us a year to get into production on the real side, simply because you have a big team that has to change the way they work to adapt customize the new dashboard analytics.”
Green adds, “Splunk allows us to iterate very quickly without needing a room full of super experts. And I think some of the speed at which we develop with Splunk flows back into the Formula 1 team. They see a little of what we’re doing and they’re like, ‘Okay, that’s pretty fast.’”
Being able to experiment so much on the esports side, given that real-world car performance parameters are so similar, has inevitably opened the door to seeing the role AI can play in strategy decisions.
Green cites the AlphaGo computer program developed by Google subsidiary DeepMind, which could beat a human at the game of Go, as an example of how the right kind of AI can outsmart the human brain.
“It’s interesting what happened at Go,” he says. “How far can we push things like that in sport? It could be a real inspiration and I personally would love to see an AI driven strategy one day.”
For now, Splunk believes the technology is just at a stage where it can help with decision-making, rather than making the final decision itself: “What we’re doing is looking at the likelihood that someone pits because of his performance,” adds Hodge.
“You can see a deterioration in lap times so that will probably be the pit window. And that’s where we’re at: we’re not quite there yet where we can race and predict everything. But we definitely help with the decision making.”
Green agrees that letting the AI take control of the show isn’t realistic right now: “Are we ready today? No. Do I think we’ll ever have an AI deciding when to get in the box? Well, there are times when you can tell it’s time to change tires because there’s this dashed yellow line [on the race corner] that tells us the pit window.”
“But there are so many parameters. You have 20 riders on the grid, you have all the different variations, you have people’s riding styles and people don’t perform and behave like you expect them to sometimes.”
“The drivers are very good but they will change their lines, there is a competitive element. So I think we need to know more about all the parameters that we can collect to really understand them.”
“And if the artificial intelligence tells you that you’re winning the race, but we want to change tires with four laps to go, would you trust it enough to follow it? Who knows? But I don’t think we’re quite there yet.”
There is another critical factor at play here: Formula 1 should remain a sport for entertainment purposes rather than a purely technical exercise. Because of this, drivers are required to drive the car “alone and unaided” and the use of automated systems to assist them has been strictly prohibited.
Perhaps such restrictions should also be imposed on pit wall decisions as part of the charm of working as a team in Formula 1 is that people sometimes make mistakes – and that helps make things unpredictable.
As Hodge says, “I play racing games. I’m not very good at it, but I’m happy. When I compete against an artificial intelligence, you are never completely satisfied. It’s not that dramatic: I defeated a computer.”
“I enjoy competing against 19 other people that I’ve never met because there’s a human element to it. There is a sport behind it. That’s why I don’t think we’ll ever reach full AI. It’s one of the reasons why in Formula 1 you still need Lando [Norris] or Oscar [Piastri] to push buttons.”
“You still need the athletic elements and a certain level of dexterity to bring the drama, the theatrics, or the heroes and villains into it,” says Hodge.
This article was written by Jonathan Noble
The original of this post “Can artificial intelligence take over Formula 1 racing strategy?” comes from Motorsport-Total.com.
