The dangerous AI tool that could dominate F1

'Alexa, can you design me a Formula 1 car?'

Whether it's getting the latest news, listening to music or asking a question, these days many people have small pieces of basic artificial intelligence in their homes.

While this might not actually be the cutting edge of AI technology, the tech industry is advancing to the point where computers could be able to think for themselves and, potentially, come up with the designs for future F1 cars.

Machine learning is the art of a computer in effect teaching itself how to accomplish tasks rather than being programmed to react in a certain way. This technology is now viewed as a potential goldmine for F1 teams.

But while it is exciting for technology buffs, it will cause headaches for the championship's bosses when it comes to trying to regulate it. It could, as one senior figure has suggested, signify a "dangerous" direction in terms of machines ultimately replacing humans.

One of the companies at the forefront of machine learning is F1 sponsor Acronis. Its boss, Serguei Beloussov, says the technology could be a gamechanger in key aspects of grand prix racing.

"In F1, there are ultimately three areas that you can apply machine learning - one is the race strategy," he says. "There is some advantage, but not so much - because a race is a highly random activity, so it is relatively difficult to make a sustainable project because there is a lot of randomisation.

"Secondly, you can apply machine learning to logistics and operations. There are a lot of logistics in F1 and it is expensive for the teams to do it, so you can maybe do better with deep learning. But mostly what teams want to do is in the third area: car design.

"Today, when you have a team like Williams with 110 aero engineers, and a top team has maybe 400-500, even the smaller team knows at the beginning of the season that the car is broken, it takes five-to-nine months of a single cycle for the 110 engineers to come up with a new design, test it and put it on the car. If it works it works, but there is no guarantee it will work. For Mercedes, it takes between one-and-a-half-to-three months, so Mercedes can fix things much faster.

"It is quite possible that we will have very dangerous developments in this area" Serguei Beloussov

"With machine learning, technically you can have a machine designing the car. You can do that with many fewer people and much faster. It can basically get the new car, test it, and after one-and-a-half months you have a better car. That is what is happening. But it has not yet happened yet."

Like any new technology that has the chance to produce a decent advantage, there is little motivation in teams going public in revealing exactly what they are up to - or how advanced their machine learning projects have become.

After all, in a hugely competitive arena such as F1, if one outfit can unlock something that proves to be a game changer, it has a powerful incentive to keep it quiet.

Before Christmas, however, Mercedes boss Toto Wolff suggested that the push towards using machine learning tools was a priority for the top teams.

"For us, and many industries, with digitalisation the environment is changing," he said at a conference for team sponsor HP. "Machine learning, artificial intelligence and robots are going to take over a lot of the work that takes a long time to process."

In the F1 world, we are not yet at the point where humans are being replaced by computers in the car design process. But the technology is moving so fast that it is highly likely that in the near future a lot of the legwork will be performed exclusively by artificial intelligence.

Pushed on how far teams have got with the technology right now, Beloussov says: "It is kind of confidential, teams are very sensitive about it because this is the most promising area right now.

"They are always limited by the rules about what they can do, but this is one area where they are not limited. So, every team is doing something. The important thing for us is that they need to keep the data, and for deep learning to learn it needs the data. It cannot learn without data.

"Until two years ago, I was certain that it would be difficult to make a machine that would be more sophisticated than a human, and that we don't know how to make such a machine. But in the past two years there has been such progress that I am no longer sure. It is quite possible that we will have very dangerous developments in this area."

His use of the word dangerous refers to the fact that the machine learning technology could become so advanced that it negates the need for human involvement. And as F1 heads towards a budget cap era where there will be a squeeze on personnel, it could set up disruptive dynamics around where teams devote resources.

"All the teams are doing it already," adds Beloussov. "It will take some time but they are doing it. My opinion is that it will play a critical role in five years and will be most important in 10. It is not going to be very long.

"There are a lot of things they do that are really repetitive activities that can be replaced by machines, who will do it much cheaper, faster and much easier. The fact [humans] do it now is just an accident."

Some might fear that F1 could become a prototype for the kind of scenarios that routinely play out in dystopian sci-fi moves. But Wolff still believes that so long as drivers continue to be central to motorsport, then the human element will remain.

Machine learning will soon to be an essential tool to winning in F1

"We have an interaction of technology with the human," he said. "An exceptional human drives the car and sometimes he will perceive things that happen in the car differently to the what the data says.

"So the trick we need to manage is that the very bright engineer, who has the data, needs to somehow find a way to compromise and merge his thinking with the feeling that Lewis [Hamilton] or Valtteri [Bottas] has in the car, and that might differ sometimes.

"On strategy, we have this bright team of strategists who have 10,000 simulations in the background and, in the end, it is a human who makes the decision. I would like to steer it a little bit away from the humans. I want more tools with faster processing and the ability to give us more suggestions and proposals.

"But then the man on the pitwall needs to take the final decision and add common sense, because racing is still common sense. It is humans racing each other, not robots. This is what we need to achieve. This is where we can gain: we are ramping up our artificial intelligence and machine learning capabilities faster than the other teams."

Currently, it really is not a case of if F1 progress becomes dominated by artificial intelligence, but when. As Beloussov points out, machine learning tools are already rampant in other sports.

"Football strategy is done by computer and the teams that don't use computers lose," he says. "The computer can figure out that this guy has a 92% chance of scoring from the right and 42% chance of scoring from the left.

"So they will tell the players to block him from the right but don't block him from the left, because he doesn't have a chance. Teams play with this kind of strategy and they win."

Machine learning will soon to be an essential tool to winning in F1 too.