Artificial intelligence: How Google is attacking the car industry
Ten years ago, Apple released its first iPhone. Today, the digital revolution is entering the next phase: new operating systems based on machine learning could make autonomous driving mainstream.
“Stay hungry. Stay foolish. Be different.” With his approach to business, Steve Jobs became the embodiment of innovation. Apple’s iPhone did not only change the way we make calls – but also how we establish new business models. The operating system mobile iOS formed the basis for apps. Most of those applications were not written by Apple themselves, but by the worldwide community of developers. With every new business idea, Apple made money – and in the process became the world’s most valuable company. Subsequently, Google created their own Android ecosystem for apps – and has since become the global market leader.
What does this all have to do with autonomous driving, you may ask? As CES and NAIAS 2017 demonstrated: a lot. In Detroit of all cities, Waymo (Google’s self-driving car company) CEO John Krafcik challenged the world’s leading automotive suppliers: “We are producing the entire suite of sensors and have been able to cut the cost of laser scanners by 90 percent,” he announced.
Google’s developers draw on ten years and around 870,000 miles (1.4 million kilometers) of experience in self-driving mobility. This fact increases the significance of Krafcik’s statement. So far, there had been a common understanding in the industry that the first driverless cars would be extremely expensive. Now Krafcik is entering the field with a Chrysler Pacifica. An affordable vehicle with a clear message: the times of basic research are over. Autonomous driving is becoming mainstream. Waymo is democratizing the technology before it is even available.
An operating system for driverless cars
To the established OEMs, this is a declaration of war – and an entirely new way to create vehicles. Modern premium cars are already complex, software-based devices with up to 100 million lines of code. But their electronics architecture is rather old school. There are 50 to 100 different central processing units (CPUs) – most of them limited to just a few functions. What they lack is a general operating system that introduces an abstraction layer between the hardware and the functions.
That would make it feasible to regularly update cars via software updates – regardless of the vehicle’s brand: “We don’t have the strength to solve all problems on our own,” says BMW board member Klaus Fröhlich. “That is why we are interested in sharing the tasks and reaching industry standardization for autonomous driving as soon as possible.” Fröhlich clearly doesn’t want to waste any time: “Tech companies like Uber are increasing the pressure to use robocabs for ride sharing as soon as possible. We have to prepare for that.”
What is still missing is central intelligence for cars that can react to unexpected events. Just like experienced human drivers, highly automated vehicles need to reliably assess what is happening on the road. “Image recognition is not sufficient,” explains Uwe Franke, Head of Image Understanding at Daimler. “The system must be able to understand the semantics of the scene to predict the outcome.” Conventional machines need preprogrammed patterns to react. But for the proactive behavior of the vehicle, that will not work. The number of possible variations is significantly higher in road traffic than in rule-based games like chess – because there is an almost unlimited number of participants. And they don’t necessarily always follow the rules.
Machine learning: The first business models arrive
To solve these problems, the cars need a new operating system that is based on machine learning. A look into the IT world shows what’s at scope for established car manufacturers: Google Search is the central hub of the web. Android is the dominant smartphone operating system, with Apple iOS being the only remaining serious contender. The market is basically cemented, the threshold for competitors with new operating systems is extremely high. “We are experiencing the same phenomenon with artificial intelligence,” explains Arwed Niestroj who runs Daimler’s Silicon Valley research center. “At the moment, we are seeing the first business models such as the natural language assistants Google Home or Amazon Alexa. The demand is so high that Facebook and others are also now developing intelligent systems.”
The digital party is entering the next phase. And everybody is trying to build their own ecosystem around artificial intelligence. “The more promising a platform seems, the more attractive it is for creatives,” says Niestroj. Programmers who have so far worked on apps or computer games are supposed to work on autonomous applications. IT giants like Alphabet (Google), Microsoft or Nvidia are paving the way towards artificial intelligence with software tools. “Nvidia doesn’t only offer high performance chips for neural networks, but also the developer environment for such programs,” explains Niestroj. “Google and others also provide affordable and virtually endless computing capacity for machine learning.” In a nutshell: the entry barriers are low and the business opportunities tempting.
Whether it’s cars or (flying) robots – start-up communities worldwide are busy experimenting with all kinds of applications. Even houses and cities will be able to manage themselves intelligently. Between these systems there are several connecting factors. It’s not only about the car, which is the most complex mobile device: it’s about the entire Internet of Things.
The logic for autonomous applications and natural language interactions between man and machine can easily be transferred from one product to the other. “It is really exciting to see how incredibly fast this area is developing,” says Niestroj. Within the next decade, intelligent operating systems are supposed to become standard. That includes autonomous vehicles. Waymo want to deliver their platform to more OEMs besides Fiat and Honda. That’s why it’s no surprise that BMW board member Klaus Fröhlich is in a rush to counter this attack with BMW’s own alliance.
Are traditional OEMs and suppliers falling behind Silicon Valley players like Google? How big is the threat for their business model? Share your thoughts in the comment section!