- John Mahoney
Driving Ambition
Carmakers around the globe have fallen for AI — but experts say it’s not nearly as clever as we think. By John Mahoney
The atmosphere in Maranello, Italy, couldn’t have been more congenial. After all, it’s not every day Ferrari reveals a sleek, new V12-powered supercar. The silk drapes were swept away, revealing the 12Cilindri, a two-seater front-engine coupé. The waiting media responded with audible gasps, even a short burst of applause.
The smile on the face of Andrea Militello, Ferrari’s head of exterior design for GT cars, said it all: pride, relief and genuine happiness that after four long years, his work was done.
“So Andrea, was the new 12Cilindri designed by AI ... or not?” one of the invited journalists asked.
Militello was initially too shocked to respond. As he gathered his thoughts, the smile remained fixed, becoming more of a grimace. His eyes narrowed.
“No, no, no! What would AI know about true design, beauty, proportions? Nothing, no. For a Ferrari, there must be emotion,” he said.
“So you don’t use AI at all?”
The second question came from an older, more seasoned journalist wearing a sports jacket two sizes too big.
“Well, of course we do, everyone does. But only to build abstract shapes, to provide an optimum form best for aero. But it’s just the beginning of the process,” Militello said.
And there it was: proof that artificial intelligence is already embedded in Ferrari’s most sacred design division.
AI has become a buzzword in the car industry, attracting a mixture of excitement and curiosity, as well as ignorance and fear.
When one thinks of AI’s use in cars, driverless vehicles are often the first example to come to mind. But many legacy carmakers are reluctant to reveal when we’ll actually be able to buy a fully autonomous vehicle.
According to Gary Marcus, a professor of psychology and neural science at New York University and an AI expert, finessing the technology has proved arduous. “It was easy to get a demo that was 80 per cent of the way there, but getting the last 20 per cent has proven really hard,” he said in a recent BBC interview. “Elon Musk promised us driverless cars that would go across the States and a million robotaxis, but we’re not seeing that yet, and he’s promised that every year from 2015 to the present.”
Those involved in developing driverless cars say the problems with developing them are almost infinite. The tech struggles with new obstacles, handwritten signs, weather changes, flashing red and blue lights and simply recognising everyday objects — and when an obstacle is a small child running out in the road wearing a dark rain jacket flapping in the wind, the outcome is almost unthinkable.
When it works, of course, it’s magical. In the meantime, while the multibillion-dollar driverless car industry irons out the final 20 per cent, AI is making inroads improving the cars we all drive.
Witness what’s happening at Germany’s Porsche. While other brands within the Volkswagen Group plough billions into driverless tech, Porsche has decided to employ the tech elsewhere. “It’s actually hard to think where we haven’t deployed AI,” says Silke Dongus-Sattler, the marque’s data and AI strategy lead. “It’s being used in all areas, covering the whole lifecycle of our vehicles, from development and production to even predicting how car parts will wear once in the hands of owners.”
She adds that in the future Porsches equipped with this technology will be capable of detecting issues long before the owner does behind the wheel. Better still, in the age of connectivity, not only will the faulty part be diagnosed before any harm has taken place, but the car can also self-order the part and even suggest available dates for the owner to drop off their vehicle at their local dealer. It sounds far-fetched, but this technology could be in our cars in a couple of years.
AI is already in use in production, inspecting paint finish on the assembly line with an accuracy far higher than any human is capable of. AI-driven software has also reconfigured production lines to cut waste and lower the risk of accidents. The technology has even been introduced in canteens: smart ordering of food at Porsche’s HQ has both reduced waste and cost.
“How can you not be an advocate of this technology?” asks Dongus-Sattler. “AI helps us create and innovate in our cars, optimise our company processes and find solutions to provide more convenience to our customers.”
Of course, the tech is not perfect — something the risk-averse marques are already aware of. As recently as 2014, AI technology gave wrong answers to questions 50 per cent of the time. Today, AI is capable of making both dumb mistakes or even weird “hallucinations” when replying.
With big brands such as Volkswagen and Mercedes embedding the tech in the infotainment, AI is about to have a very public face. The rewards could be huge, but so could the pitfalls — especially in litigious markets like the US.
Mercedes-Benz’s chief technology officer, Markus Schäfer, is well aware of the dangers involved with rolling out AI in its cars. “It’s not something you implement in a car and then just leave it,” said Schäfer in an interview with Car magazine. “If you sit in a car and ChatGPT tells you something that’s absolute nonsense, [the maker] might be exposed to product liability cases.”
This has prompted Benz to put a temporary leash on just what the tech can do, and has forced the carmaker to accept liability for any accident that occurs when its latest S-Class is in driverless mode.
Thanks to embedded ChatGPT, in the future your car will be able to learn your routines, daily habits and be capable of reading both your emails and calendar. It will remind you of birthdays, anniversaries and can suggest the best Italian in town after it’s read all the reviews. It will soon be able to do all manner of menial tasks, including replying to emails in your writing style.
Some carmakers hint that one day you’ll even be able to have a casual chat or have the tech be there when you pour your heart out following a break-up — the perfect witness to your ugly crying face. But this prospect of people falling in love with their cars has experts such as NYU’s Marcus worried.
Despite what you’ve seen in the movies, AI is poor at both logic and reasoning, and struggles to be emotionally intelligent. Any display of empathy, then, is fleeting. Even the most advanced avatar normally fails when it comes to possessing the imperfect, raw emotions needed for it to be considered a real human. And until it learns them, according to some, AI will never be capable of designing Ferraris.
This is an extract from an article that appears in print in our twelfth edition, Page 34 of Winning Magazine with the headline: “Driving Ambition”. Subscribe to Winning Magazine today.
コメント