“Where would you like to go?” Siri asked.
It was a sunny, slightly dreamy morning in the heart of Silicon Valley, and I was sitting in the passenger seat of what seemed like a perfectly ordinary new car. There was something strangely Apple-like about it, though. There was no mistaking the apps arranged across the console screen, nor the deadpan voice of Apple’s virtual assistant, who, as backseat drivers go, was pretty helpful. Summoned via a button on the steering wheel and asked to find sushi nearby, Siri read off the names of a few restaurants in the area, waited for me to pick one, and then showed the way on a map that appeared on the screen.
The vehicle was, in fact, a Hyundai Sonata. The Apple-like interface was coming from an iPhone connected by a cable. Most carmakers have agreed to support software from Apple called CarPlay, as well as a competing product from Google, called Android Auto, in part to address a troubling trend: according to research from the National Safety Council, a nonprofit group, more than 25 percent of road accidents are a result of a driver’s fiddling with a phone. Hyundai’s car, which goes on sale this summer, will be one of the first to support CarPlay, and the carmaker had made the Sonata available so I could see how the software works.
CarPlay certainly seemed more intuitive and less distracting than fiddling with a smartphone behind the wheel. Siri felt like a better way to send texts, place calls, or find directions. The system has obvious limitations: if a phone loses the signal or its battery dies, for example, it will stop working fully. And Siri can’t always be relied upon to hear you correctly. Still, I would’ve gladly used CarPlay in the rental car I’d picked up at the San Francisco airport: a 2013 Volkswagen Jetta. There was little inside besides an air-conditioning unit and a radio. The one technological luxury, ironically, was a 30-pin cable for an outdated iPhone. To use my smartphone for navigation, I needed a suction mount, an adapter for charging through the cigarette lighter, and good eyesight. More than once as I drove around, my iPhone came unstuck from the windshield and bounced under the passenger seat.
Android Auto also seemed like a huge improvement. When a Google product manager, Daniel Holle, took me for a ride in another Hyundai Sonata, he plugged his Nexus smartphone into the car and the touch screen was immediately taken over by Google Now, a kind of super-app that provides recommendations based on your location, your Web searches, your Gmail messages, and so on. In our case it was showing directions to a Starbucks because Holle had searched for coffee just before leaving his office. Had a ticket for an upcoming flight been in his in-box, Holle explained, Google Now would’ve automatically shown directions to the airport. “A big part of why we’re doing it is driver safety,” he said. “But there’s also this huge opportunity for digital experience in the car. This is a smart driving assistant.”
CarPlay and Android Auto not only give Apple and Google a foothold in the automobile but may signal the start of a more significant effort by these companies to reinvent the car. If they could tap into the many different computers that control car systems, they could use their software expertise to reimagine functions such as steering or collision avoidance. They could create operating systems for cars.
Google has already built its own self-driving cars, using a combination of advanced sensors, mapping data, and clever navigation and control software. There are indications that Apple is now working on a car too: though the company won’t comment on what it terms “rumors and speculation,” it is hiring dozens of people with expertise in automotive design, engineering, and strategy. Vans that belong to Apple, fitted with sensors that might be useful for automated driving, have been spotted cruising around California.
After talking to numerous people with knowledge of the car industry, I believe an Apple car is entirely plausible. But it almost doesn’t matter. The much bigger opportunity for Apple and Google will be in developing software that will add new capabilities to any car: not just automated driving but also advanced diagnostics and over-the-air software upgrades and repairs. Already, a button at the bottom of the Android Auto interface is meant for future apps that could show vehicle diagnostics. Google expects these apps to be made by carmakers at first, showing more advanced vehicle data than the mysterious engine light that flashes when something goes wrong. Google would like to make use of such car data too, Holle says. Perhaps if Android Auto knew that your engine was overheating, Google Now could plan a trip to a nearby mechanic for you.
At least for now, though, the Google and Apple services essentially can read only basic vehicle data like whether a car is in drive, park, or reverse. Carmakers won’t let those companies put their software deeper into the brains of the car, and whether that will change is a crucial question. After all, modern cars depend on computers to run just about everything, from the entertainment console to the engine pistons, and whoever supplies the software for these systems will shape automotive innovation. Instead of letting Apple and Google define their future, carmakers are opening or expanding labs in Silicon Valley in an attempt to fend off the competition and more fully embrace the possibilities offered by software.
The car could be on the verge of its biggest reinvention yet—but can carmakers do it themselves? Or will they give up the keys?
Cars are far more computerized than they might seem. Automakers began using integrated circuits to monitor and control basic engine functions in the late 1970s; computerization accelerated in the 1980s as regulations on fuel efficiency and emissions were put in place, requiring even better engine control. In 1982, for instance, computers began taking full control of the automatic transmission in some models.
New cars now have between 50 and 100 computers and run millions of lines of code. An internal network connects these computers, allowing a mechanic or dealer to assess a car’s health through a diagnostic port just below the steering wheel. Some carmakers diagnose problems with vehicles remotely, through a wireless link, and it’s possible to plug a gadget into your car’s diagnostic port to identify engine problems or track driving habits via a smartphone app.
However, until now we haven’t seen software make significant use of all these computer systems. There is no common operating system. Given that carmakers are preventing CarPlay or Android Auto from playing that role, it’s clear that the auto companies are taking a first crack at it. How successful they are will depend on how ambitious and creative they are. Roughly 10 minutes north of Google’s office, I got to see how one of the oldest car companies is beginning to think about this possibility.
Ford opened its research lab in Palo Alto in January. Located one door down from Skype and just around the corner from Hewlett-Packard, it looks like a typical startup space. There are red beanbags, 3-D printers, and rows of empty desks, which the company hopes to fill with more than a hundred engineers. I met a user-interface designer named Casey Feldman. He was perched atop a balance board at a standing desk, working on Ford’s latest infotainment system, Sync 3. It runs software Ford has developed, but the automaker is working on ways to hand the screen over to CarPlay or Android Auto if you plug in a smartphone. Feldman was using a box about the size of a mini-fridge, with a touch screen and dashboard controls, to test the software. He showed how Sync 3 displays a simplified interface when the car is traveling at high speed.
Ford’s first touch-screen interface, called MyFord Touch, didn’t go well. Introduced in 2010, it was plagued by bugs, and customers complained that it was overcomplicated. When Ford dropped from 10th to 20th place in Consumer Reports’ annual reliability ratings in 2011, MyFord Touch was cited as a key problem. The company ended up sending out more than 250,000 memory sticks containing software fixes for customers to upload to their cars.
Besides running apps like Spotify and Pandora Radio, Sync 3 can connect to a home Wi-Fi network to receive bug fixes and updates for the console software. Ford clearly hopes that drivers will prefer its system to either CarPlay or Android Auto, and it’s doing its best to make it compelling. “It’s a cultural shift,” says Dragos Maciuca, the lab’s technical director. The lab wants to incorporate “some of the Silicon Valley attitudes, but also processes” into the automotive industry, he says. “That is clearly going to be very challenging, but that’s why we’re here. It doesn’t make sense that you buy a car, and the first thing you do is buy a $5 suction cup for your phone.”
Ford has been ahead of many automakers in its experimentation. It has come out with a module known as Open XC, which lets people download a wide range of sensor data from their cars and develop apps to aid their driving. A Ford engineer used it to create a shift knob for cars with manual transmission so that the stick lights up or buzzes when it’s time to change gears. But Open XC has not taken off widely, and despite Ford’s best efforts, the company’s overall approach still seems somewhat conservative. Maciuca and others said they were wary of alienating Ford’s vast and diverse customer base.
In February, meanwhile, the chip maker Nvidia announced two new products designed to give cars considerably more computing power. One is capable of rendering 3-D graphics on up to three different in-car displays at once. The other can collect and process data from up to 12 cameras around a car, and it features machine-learning software that can help collision-avoidance systems or even automated driving systems recognize certain obstacles on the road. These two systems point to the huge opportunity that advanced automotive sensors and computer systems offer to software makers. “We’re arguing now you need supercomputing in the car,” Danny Shapiro, senior director of automotive at Nvidia, told me.
If anyone could find a great use for a supercomputer on wheels, it’s Chris Gerdes, a professor of mechanical engineering who leads Stanford University’s Dynamic Design Lab. Gerdes originally studied robotics as a graduate student, but while pursuing a PhD at Berkeley, he became interested in cars after rebuilding the engine of an old Chevy Cavalier. He drove me to the lab from his office in an incredibly messy Subaru Legacy.
Inside the lab, students were working away on several projects spread across large open spaces: a lightweight, solar-powered car; a Ford Fusion covered in sensors; and a hand-built two-person vehicle resembling a dune buggy. Gerdes pointed to the Fusion. After Ford gave his students a custom software interface, they found it relatively easy to get the car to drive itself. Indeed, the ability to manipulate a car through software explains why many cars can already park themselves and automatically stay within a lane and maintain a safe distance from the vehicle ahead. In the coming years, several carmakers will introduce vehicles capable of driving themselves on highways for long periods. “There are so many things you can do now to innovate that don’t necessarily require that you bend sheet metal,” Gerdes said as we walked around. “The car is a platform for all sorts of things, and many of those things can be tried in software.”
The dune-buggy-like car takes programmability to the extreme. Virtually every component is controlled by an actuator connected to a computer. This means that software can configure each wheel to behave in a way that makes an ordinary road feel as if it were covered with ice. Or, using data from sensors fitted to the front of the car, it can be configured to help a novice motorist react like a race-car driver. The idea is to explore how computers could make driving safer and more efficient without taking control away from the driver completely.
In fact, one small carmaker—headquartered in Silicon Valley—shows how transformational an aggressive approach to software innovation could be.
Tesla Motors, based in Palo Alto, has built what’s probably the world’s most computerized consumer car. The Model S, an electric sedan released in 2012, has a 17-inch touch-screen display, a 3G cellular connection, and even a Web browser. The touch screen shows entertainment apps, a map with nearby charging stations, and details about the car’s battery. But it can also be used to customize all sorts of vehicle settings, including those governing the suspension and the acceleration mode (depending on the model, it goes from “normal” to “sport” or from “sport” to “insane”).
Every few months, Tesla owners receive a software update that adds new functions to their vehicle. Since the Model S was released, these have included more detailed maps, better acceleration, a hill-start mode that stops the car from rolling backwards, and a blind-spot warning (providing a car has the right sensors). Tesla’s CEO, Elon Musk, has said a software patch released this summer would add automated highway driving to suitably equipped models.
These software updates can do more than just add new bells and whistles. Toward the end of 2013, the company faced a safety scare when several Model S cars caught fire after running over debris that ruptured their battery packs. Tesla engineers believed the fires to be rare events, and they knew of a simple fix, but it meant raising the suspension on every Model S on the road. Instead of requiring owners to bring their cars to a mechanic, Tesla released a patch over the airwaves that adjusted the suspension to keep the Model S elevated at higher speeds, greatly reducing the chance of further accidents. (In case customers wanted even more peace of mind, the company also offered a titanium shield that mechanics could install.)
Tesla’s efforts show how making cars more fully programmable can add value well after they roll out of the showroom. But software-defined vehicles could also become a juicy target for troublemakers.
In 2013, at the DEF CON conference in Las Vegas, two computer-security experts, Charlie Miller and Chris Valasek, showed that they could hijack the internal network of a 2010 Toyota Prius and remotely control critical features, including steering and braking. “No one really knows a lot about car security, or what it’s all about, because there hasn’t been a lot of research,” Miller told me. “It’s possible, if you went out and bought a 2013, they’ve done huge improvements—we don’t know. That’s one of the scary things about it.”
A few real-world incidents point to why car security might become a problem. In February 2010, dozens of cars around Texas suddenly refused to start and also, inexplicably, began sounding their horns. The cars had been fitted with devices that let the company that leased them, the Texas Auto Center, track them and then disable and recover them should the driver fail to make payments. Unfortunately, a disgruntled ex-employee with access to the company’s system was using those gadgets to cause havoc.
I asked Gerdes whether concerns over reliability and security could slow the computerization of cars. He said that didn’t have to be the case. “The key question is, ‘How fast can you move safely?’” he says. “The bet that many Silicon Valley companies are making—and that many car companies are making with their Valley offices—is that there are ways to move faster and still be safe.”
Ultimately, the opportunities may well outweigh such concerns. Tesla’s efforts point to how significant software innovation could turn out to be for carmakers. Tesla is even experimenting with connecting the forthcoming autopilot system to the car’s calendar, for example. The car could automatically pull up outside the front door just in time for the owner to drive to an upcoming appointment.
Perhaps this also explains why Apple and Google are now dabbling in vehicle hardware: so they can fully own some people’s driving time even before carmakers decide to open up more aspects of their vehicles. “Clearly Apple and Google would love to be the ones who have the operating system for these future cars,” Gerdes says.
As I drove back to the San Francisco airport, my VW Jetta felt more low-tech than ever. The ride was fairly peaceful, with the Santa Cruz Mountains looming in the distance. Even so, after so much driving, I would’ve been glad had Siri offered to take over.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.