Just a few months ago, Apple was believed to have given up hopes, albeit temporarily, of making its own autonomous vehicle. Of course, the company has neither confirmed nor denied it was even working on one, but it seems it never really took that off the table in the first place. At a machine learning conference called NIPS, Apple director of AI research Ruslan Salakhutdinov revealed some of the projects Apple has going internally. Curiously, almost all of them can be connected to systems needed for a self-driving car.
Rumors of Apple making its own self-driving car reach its peak around the time of Google’s own project. As the hype around autonomous vehicles died down, so did Apple’s interest. Or so it seemed.
At NIPS, Salakhutdinov talked specifically about four machine learning projects that were all too obviously related to self-driving cars. One project, which has already been posted in a research paper online, showed how software was trained using lidars to identify pedestrians and cyclists. Lidars are one of the sensors essential in self-driving cars.
Three other projects were mentioned but have not yet been published. One related to being able to identify cars, pedestrians, roads, and more with a camera or multiple cameras, even when the lenses have been spattered by raindrops or when pedestrians were obscured by parked cars. Another project used simultaneous localization and mapping or SLAM to give robots, and cars, a sense of direction. The fourth project used sensors to build a 3D map of the car’s surrounding, including traffic lights and road signs.
Granted, neither Salakhutdinov nor Apple at large talked specifically about using these lessons in making self-driving cars. But while it may have been simply presenting its own machine learning and AI developments, it might have also been trying to scout for new talent that could help finally push its automotive ambitions out of the woodwork.