Companies scramble to distance themselves from Uber after massive self-driving setback.
The International Consumer Electronics Show (CES), held January in Las Vegas every year, draws almost 200,000 visitors making it the place for every tech vendor to trot out their latest gee-whiz technology. In that circus-like environment, even big companies have to compete for attention from the over 7,000 media invited.
At CES 2018, there was NVIDIA, known by most of the audience for its high-end gaming cards. But NVIDIA was getting into some serious business, eager to show the world that it was not all about gaming. Its new hardware could help AI researchers and the development self-driving cars. It was there that CEO Jensen Huang announced what sounded like a technology partnership with Uber.
About two months later, Uber is involved in a fatal accident, running over a pedestrian with its self-driving car. The accident sends shock waves through the electronics, computing and automotive industries—as well as society at large. Company execs scramble to distance themselves from Uber, their technologies from whatever Uber was using.
No doubt, Mr. Huang wishes he had picked another announcement to make.
Partner or Customer?
After the fateful accident, the NVIDIA “partnership” with Uber has come under scrutiny. The media at NVIDIA’s big user event, GTC (recently concluded), abuzz with a big breaking story, interrupted the show demanding to know how involved with Uber NVIDIA had become.
Huang himself had said “the collaboration utilizes NVIDIA technology for Uber Advanced Technologies Group’s fleets of self-driving cars and freight trucks, running AI algorithms that enable vehicles to perceive the world, predict what will happen next and quickly choose the best course of action, even in complex environments.” That’s from an NVIDIA press release on January 7, 2018,
The AI algorithms mentioned above…were they Uber’s algorithms or NVIDIA’s? It will be hard to fault hardware unless it failed. Otherwise, blaming a GPU for the fatality would be like blaming the vehicle, the Volvo XC90. Both are relying on instructions, whether they come from computers or humans.
Volvo’s XC90 SUV had its own IntelliSafe collision detection and avoidance system onboard, which also should have prevented such an accident. The Volvo system had been disconnected in favor of Uber’s system on top of the vehicle, according to a spokesman from Aptiv, maker of radar and camera components used in the Volvo. It would not be unusual to disconnect one system if another one was operational or being tested. Having two systems active for collision detection and avoidance would have been redundant and possibly sent conflicting actions to the vehicle systems for steering and braking.
But the GPU technology may have been nothing more than the GPU chips themselves, not NVIDIA software, not the autonomous vehicle (AV) platforms the company offers.
“Uber does not use NVIDIA DRIVE technology. Uber develops its own sensing and drive technology,” Huang said in response to questions at GTC. It is a subdued response, not like the excited promotion of announcing a partnership. More like not wanting to speak ill of someone you broke up with.
”Uber developed their own sensing and driving applications and software,” Huang said later from the show floor, where he was featured on video on CNBC’s “Mad Money.” “Their technology and our technology is completely different,” added Huang.
NVIDIA is going to pause, wait until an investigation is performed, and learn what it can, Huang relayed, as he announced that public-road testing of NVIDIA’s own fleet of five self-driving cars was suspended.
NVIDIA makes much of its self-driving technology. Its self-driving platform, DRIVE, is used by roughly 370 of the company’s partners around the world. Again, Uber is not one of those partners.
Although, GTC was full of self-driving car technology, not just NVIDIA’s but Velodyne, makers of the Uber’s LiDAR. Uber was not on the show floor with other vendors but did have speakers present.
GPU chips are brains of most self driving car systems, but the systems also demand other hardware and the software that ties all the hardware together. Uber has a large research and development team and the company seems bent on creating its own hardware and software for its own purpose. But as Uber has $21 billion in funding, it’s not big enough to produce microprocessors. A semiconductor wafer fabrication plant is a $10 billion project. And the competition is well established. There are a few of them in the world (Intel and Samsung leaders in CPUs, NVIDIA and AMD with GPUs), so best to use, rather than invent. But as each has only recently tried to establish itself with systems that build on their chips, offering boards and platforms with AI and AV technology, it’s game on, pitting one company’s group of PhDs against the other.