Apple often maintains a ridiculously high level of secrecy around its yet-to-launch products. Back in October, it had fired an engineer who allowed his daughter to make a video of the iPhone X before the phone hit the store shelves. But the company is now allowing its researchers to share their work – primarily related to the car project – with the wider scientific community.
Speaking at the NIPS machine learning conference (via Wired), Apple’s AI director Ruslan Salakhutdinov provided some previously unpublicized details about the self-driving car technology. He offered a sneak peek into how Apple is using artificial intelligence and machine learning to detect pedestrians and make autonomous driving safer. More than 8,000 people attended the NIPS conference.
He demonstrated a system that can identify objects such as pedestrians and cars in tricky situations. It works even when raindrops have covered the camera lenses or when pedestrians and other objects are hidden behind a parked car. The system can even estimate where the pedestrians might be. “If you asked me five years ago, I would be very skeptical of saying ‘Yes, you could do that,'” Salakhutdinov said.
There has been much talk in recent years about disruption and trying to pick companies that will disrupt their industries. The debate continued at the Morningstar Investment Conference as Bill Nygren of Oakmark Funds faced off with Morgan Stanley's Dennis Lynch. Q2 2021 hedge fund letters, conferences and more Persistence Morningstar's Katie Reichart moderated the Read More
How it will handle urgent situations
He discussed how the tech giant is collecting data from vehicles on the road to build detailed 3D maps to let the future cars navigate the unknown streets. It will offer details such as road markings and traffic lights. Such maps will play an important role in the operation of autonomous vehicles. Salakhutdinov also offered details about providing directions to the cars through simultaneous localization and mapping using various sensors and dynamic decision-making in urgent situations such as a wayward pedestrian.
His talk comes just weeks after two Apple researchers Yin Zhou and Oncel Tuzel published a research paper in arXiv to discuss a new technology to detect objects on the road more accurately. Salakhutdinov also discussed the same technology at the NIPS conference. Most self-driving cars rely on LiDAR technology to detect 3D objects. However, LiDAR can’t detect small objects that are far away.
Zhou and Tuzel had proposed a new technology called VoxelNet that uses machine learning and artificial intelligence to identify pedestrians, cyclists, and other objects on the road. In computer simulations, VoxelNet performed way better than LiDAR in detecting 3D objects. But the technology is yet to be tested on the roads.
Apple loosening the grip on secrecy
The publication of a research paper followed by Salakhutdinov’s talk suggests that Apple is keen to show its AI research in the field of autonomous driving to the wider world. Or it could simply be that the company is trying to attract talent. The tech giant had been struggling to attract AI talent because most scientists prefer academic and industry recognition. Apple’s secretive approach might have turned them off. That’s why Apple launched its own Apple Machine Learning Journal earlier this year.
Apple has already been testing its car project on public roads in California for months using Lexus vehicles. Recent reports suggest that the company was planning to test its autonomous driving technologies away from the public eye. Apple has reportedly leased the proving grounds in Surprise, Arizona that was owned by Chrysler in the past. The facility is spread over nearly 5,000 acres.
According to Jalopnik, the proving grounds include a high-speed oval, a city traffic course, steep grades, and areas for wet weather testing. Chrysler had originally set up the facility to test the impact of high temperatures on cars and various components. Apple has also been hiring technicians and test engineers from other proving grounds in the state for its own autonomous car project.
How will Apple commercialize the self-driving technology?
Apple was originally rumored to be developing full-fledged autonomous vehicles. But the project hit several roadblocks, and the company eventually shifted the focus of its car project to develop the autonomous driving systems. Tim Cook said earlier this year that self-driving technology was the “mother of all AI projects.” Apple is developing the autonomous driving systems from a core technology point of view.
Earlier this year, the iPhone maker also sought changes in the California DMV’s self-driving car testing policy. The company argued that the current rules would not help the public accept autonomous cars as safe. It is still not clear when and how Apple will commercialize its autonomous driving technology. It could license the technology to other automakers or build its own cars in the future – the latter being more likely.
For now, Apple is planning to start a shuttle service between different buildings of its campus. The shuttle vehicles will use its driverless car technology.