In its various forms, artificial intelligence (AI) is increasingly relevant in contemporary industry, including motorsport. There is no shortage of understandable fears among public opinion that without appropriate regulation and planning there is a risk of approaching independent machines with autonomous decision-making capabilities, with consequences that are not easily predictable. The second concern is that artificial intelligence and machine learning can relegate humans to a marginal, almost superfluous role. However, humanity can and must continue to carry out one of the central activities of its history: ask and ask yourself the right questionsan approach also valid for Formula 1.
The jokester
There is a striking resemblance between the technologies of our century and those predicted by Isaac Asimov in the 1956 story “The Joke Teller.” Nearly seventy years in advance, Asimov imagines a supercomputer named Multivac, forerunner of modern artificial intelligence, with such complexity that it requires a huge crowd of people to control it. The figure considered to be of the highest rank is that of the Grand Masters, individuals with very rare characteristics whose greatest task is to think of the right questions to ask the computer.
“From the beginning in Multivac’s history, it was clear that the main difficulty was how to ask questions. Multivac could solve humanity’s problems, any problem, if… if he was asked the right questions. But as knowledge progressed at an ever-faster pace, the right questions became increasingly difficult to pinpoint.” (Isaac Asimov, “The Joke Teller”). If you think about it, this is a way of working that is already current in Formula 1.
Simulate with knowledge
Take a step back from artificial intelligence and return to virtual simulations. For reasons of time, energy and computing resources, even in a technologically cutting-edge sector such as Formula 1 there are no simulations that predict every possible aspect of a physical phenomenon. Depending on the needs, the structural analyzes for example examine only the mechanical stresses or, if necessary, include the thermal ones. For CFD aerodynamic studies, however, the designer has the task of setting up a stationary or transient, or adiabatic or heat exchange simulation.
The case of Red Bull in 2012 is relevant again, where initially the blown exhausts did not work as expected, an excellent testimony to the importance of carefully choosing what to ask of a simulation. Adrian Newey says: “One explanation proposed by one of our aerodynamicists, Craig Skinner, was that this was due to the pulsations of the exhaust gases. When the cylinder valves open, a shock wave is created, a pulse that when it reaches the end of the exhaust pipe creates a ring-shaped vortex. […] Craig was able to find several studies on the topic and created a transient CFD modelapplying it to our bodywork”, (Adrian Newey, “How to build a car”).
It is only by setting the CFD to a transient rather than stationary simulation that the problem becomes apparent. Red Bull asks the analysis software to simulate a certain condition, taking into consideration the consequences of a specific physical aspect initially considered irrelevant. The dynamic also refers to that of the phenomenon of porpoising, overlooked by the FIA and teams at the dawn of the 2022 regulations. Now as then, the difficulty in finding the answer to a problem lies in asking the right question. The same is destined to happen with artificial intelligence, a tool with increasingly greater computing capabilities, as long as you know what to look for.
The logic of artificial intelligence
The potential of machine learning and artificial intelligence algorithms lies in the enormous ability to analyze already existing data samples. By evaluating every possible parameter, the software hypothesizes changes to the object of interest, predicting their effect based on previous changes. AI helps Formula 1 teams refine and optimize, starting from already known experiences. In Asimov’s story, the Grand Master’s interest is in tracing the origin of the jokes. To allow Multivac – the supercomputer – to develop an answer, he provides it with the right impulses, telling it countless jokes. It is the symbolic image of an artificial intelligence at work starting from previous experiences, a deductive approach very similar to that of humans. The main difference, however, lies in not being able to emulate the same creative, intuitive and spontaneous spirit.
If machine learning and AI had been employed in post-World War II racing, they would have looked at how manufacturers had worked up until then. Probably, the algorithms would have designed cars with the aim of reaching maximum speed on the straight, seeking the compromise between aerodynamic penetration and the dimensions of increasingly large engines. At a certain point, however, the designers of the time wondered what would happen by moving the engine from the front to the rear, a question sufficient to force a paradigm shift. It is difficult to imagine whether artificial intelligence would have asked itself the same question, since even if it wanted to experiment and proceed through attempts, these would still have to be addressed from the outside.
Ask to win
Tomorrow like yesterday, the tools change but the winning approach remains the same. Thinking about the right questions, constantly questioning what you have done up to that point. “What if it were done like this instead of there?”. Artificial intelligence presents itself as a promising tool, whose interpretation of data varies depending on what is asked of it. The possible fields of application are multiple: race strategies, set-up research, aerodynamic design and more.
Its diffusion is still slowed down by the accumulation of a sufficiently large historical data. The paradox of artificial intelligence it lies precisely in this: there is no sample large enough to include every possible case study. When a new situation arises, it is still up to the engineers to react with their creativity. Even today, in a hyper-predictive Formula 1 between computers and simulators, it happens to witness a Red Bull caught off guard on the set-up in Singapore or to see the teams experimenting directly on the track, taking note of the predictive limits of the wind tunnels .
Artificial intelligence and machine learning are preparing to transform the work of engineers in a way not dissimilar to how CFD and virtual simulations have changed what was done previously. There are two main challenges: understanding before others how to exploit the potential of emerging technologies; measure the advantages with appropriate risk control. Going against the spirit of Formula 1 would instead be to reject technological progress a priori, but this does not mean that their contraindications must still be respected and managed.
#Formula #artificial #intelligence #importance