Meet 'Professor X,' the AI genius who left his lab
Source: Matt Weinberger
In the spring of 2016, Dr. Jianxiong Xiao — affectionately known among students and staff as "Professor X" — said goodbye to his plum professorship at Princeton and his post as the founding director of the school's Computer Vision and Robotics Labs.
By the fall of that same year, Xiao, known as something of a risk-taker, had moved himself and his family from New Jersey to Silicon Valley, and raised some modest seed funding for his new startup focused on self-driving cars.
His startup, dubbed AutoX, has done its best to stay under the radar to date — apart from a filing with the California DMV to test self-driving vehicles.
The filing officially put the professor's mysterious startup in the company of giants, such as Tesla, Waymo (formerly the Google self-driving car project), Uber, and numerous other big auto companies testing self-driving cars.
But Xiao isn't worried about getting run over by the giants, saying that his small team of academics possesses the kind of expertise in computer vision that big corporations just can't match. Exhibit A: after only six months on the job, Xiao says he's already developed a prototype vehicle that can do the same things as the cars made by his deep-pocketed rivals, at a fraction of the cost.
AutoX gave a first peek at its creation on Friday, with a debut video showing its prototype system in action. The car itself isn't anything special in terms of style (it's basically just a regular 2017 Lincoln MKZ that's been rigged with AutoX technology), but it deftly navigates residential streets near San Jose, seeming to handle driving situations such as cloudy days and night-time, historically a challenge for self-driving cars, with ease.
As noted in the video, AutoX's system doesn't rely on the LIDAR laser arrays or other expensive sensors that most self-driving cars, including Waymo's and Uber's, require to function. Instead, AutoX uses advanced artificial intelligence to "see" through cameras mounted on the car and steer the car accordingly.
Better yet, Xiao tells Business Insider that the cameras that power this AutoX prototype were purchased at Best Buy for $50 a pop. "It could not be cheaper than that," he says.
From Xiao's standpoint, that's a crucial point: While future iterations of AutoX technology will support ultrasonic sensors and LIDAR and all that stuff for the sake of enhancing driver safety, the startup is currently focused on building the cheapest and most accessible system for self-driving cars that it possibly can.
To Xiao, self-driving cars have the potential to benefit society, from cutting down on traffic, to providing more autonomy for the disabled, to making long-distance trucking safer and more efficient. The next step for AutoX is building out a fleet of test cars, so it can test the technology with a variety of different vehicles types.
"Self-driving [cars] shouldn't just be a luxury, but be available to every citizen," Xiao says.
A thoughtful speaker, Xiao says he got the nickname Professor X because some of his peers at Princeton found his actual name "very complicated" (his personal web page includes a link to hear a sound of how his name is pronounced).
The term for Xiao's specific field of study is "computer vision," a branch of artificial intelligence that's just as applicable to self-driving cars as it is to Snapchat puppy filter selfies.
For Xiao, it's been a long-time area of interest. Over the years, he's won recognition from the likes of Google and Amazon for his and his team's advancements in the field of computer vision. In 2013, Xiao received his PhD from MIT, right before he went to Princeton.
The impetus for going from academia to Silicon Valley was simple, he says: He had long seen self-driving cars as a huge potential market for computer vision. After consulting with his network of computer vision and autonomous driving experts in academia, he decided the time was right to take a big leap and go into business.uber self-driving car lidarUber
It's that academic pedigree that gives AutoX an advantage, Xiao says, as it "enables us to to tap into the academic research network." With artificial intelligence experts in such high demand in Silicon Valley, it helps with recruiting that Xiao and his team already have personal relationships with many of the best and brightest in the field.
"We've known these people in person for many years," Xiao says.
Nowadays, Xiao says, AutoX is about 20 people strong, almost all engineers, with PhD-level computer vision talent that had previously worked at companies like Apple, Magic Leap, and Microsoft.
And with all that brainpower on board, Xiao says that they were able to build their prototype from scratch in only six months, without using anyone's technology. With the Waymo/Uber self-driving car IP lawsuit rocking Silicon Valley, that's a huge plus.
Xiao says the shift from academia to startup has required a subtle change in how he approaches problems. In academia, you tend to flit from one project to the next. At a private company like AutoX, you "actually make things work, actually get things done" in service of one big idea. In this case, self-driving cars.
Driven to success
While Xiao stresses this system is still a prototype, he does say it indicates the way he wants to see AutoX go, with a focus on building real self-driving technology that can handle every situation, versus more limited driver-assist features like Tesla's autopilot, which can only be engaged on freeways.
AutoX isn't much interested in manufacturing cars and going head-to-head with Tesla, Xiao says. He's equally uninterested in following the controversial Comma.ai into helping people give their existing cars limited self-driving features.
Instead, Xiao says, he's looking to partner up with auto manufacturers for their future vehicles. AutoX provides the core technology, almost like an operating system, that car companies can then take, customize to their exact needs, and use as the basis for their own autonomous systems.
Additionally, Xiao says that AutoX is similarly looking to license out its software to trucking companies, factory operators, and the like.
Xiao also wants to distinguish between AutoX and the technology from companies like Mobileye, which Intel bought for $15 billion this week — Mobileye helps self-driving car "see," sure, but other software has to take its perceptions and translate it into action for the car. AutoX is the whole package, says Xiao.
"We're building the brains for self-driving vehicles," Xiao says.