Nvidia claims autonomous driving breakthrough, but let’s see

Written by Jamie Davies
08 January 2018 @ 13:38

Nvidia has attempted to jump-start the CES PR euphoria, claiming it can achieve Level 5 autonomous driving right now with its Xavier processors.

The chip itself was initially announced 12 months ago, but this quarter has seen the processor delivered to customers. Testing has begun, and Nvidia has been stoking the fire with a very bold claim.

“Delivering the performance of a trunk full of PCs in an auto-grade form factor the size of a license plate, it’s the world’s first AI car supercomputer designed for fully autonomous Level 5 robotaxis,” Nvidia said on its blog.

Hyping up a product to almost undeliverable heights is of course nothing new in the tech world, and Nvidia has learned from the tried and tested playbook. Make an incredibly exceptional claim for a technology which is unlikely to be delivered to the real world for decades.

Xavier will form part of the Nvidia’s Drive software stack, containing 9 billion transistors. It is the product of a four-year project, sucking up $2 billion in research and development funds, with contributions from 2,000 engineers. It is built around an 8-core CPU, a 512-core Volta GPU, a deep learning accelerator, computer vision accelerators and 8K HDR video processors. All to deliver Level 5 autonomous driving.

Just as a recap, Level 5 autonomous driving is the holy grail. At this point, humans will not be needed to interact with the car at any point:

Level 0: Full-time performance by the human driver
Level 1: Driving assistance of either steering or acceleration/deceleration using information about the driving environment. Human drives the rest of the time.
Level 2: The system can be responsible for both steering and acceleration/deceleration using information about the driving environment. This could be described as hands off automation.
Level 3: This is known officially as conditional automation. The autonomous driving system will be responsible for almost all aspects of the dynamic driving task. Humans will still need to be aware to intervene in certain circumstances. This could be described as eyes off automation.
Level 4: The car will be almost fully-autonomous, though there might be rare circumstances where a human would have to intervene. Aside from the most extreme circumstances, this could be described as mind off automation.
Level 5: Full autonomy. You don’t even have to be awake.

During the same pre-CES event, the team also announced AR products, new partnerships and solutions in the gaming space, but Level 5 autonomy is the headline maker. Reaching this level is all well and good, but the technology does not have a foot in reality just yet. Nvidia might be there in terms of technological development, so it claims, but that does not mean autonomous cars will be hitting the roads any time soon. Not by a long way.

Firstly, while the processors might be there, the information is not. Companies like Google have been doing a fantastic job at creating mapping solutions, but the details is still not there for every single location on the planet. Until you can accurately map every single scenario and location a car may or may not end up in, it is impossible to state with 100% accuracy that Level 5 autonomous vehicles are achievable.

Secondly, to live the autonomous dream, a smart city is necessary. To optimize driving conditions, the car will need to receive data from the traffic lights to understand the flow of vehicles, and also any unusual circumstances. To ensure safety and performance, connectivity will have to be ubiquitous. The smart city dream is miles away, and therefore the autonomous vehicles dream is even further.

Thirdly, even if the technology is there, everything else isn’t. Regulations are not set up to support autonomous vehicles, neither is the insurance industry or the judicial system. If an autonomous vehicle is involved in a fatal incident, who get prosecuted? Do individuals need to be insured if they are asleep in the car? There are many unanswered questions.

Finally, when will we accept autonomous vehicles? Some people are incapable of sitting in a passenger seat while a loved one drives, how will these individuals react to a computer taking charge? Culturally, it might be a long time before the drivers of the world are comfortable handing control over to a faceless piece of software.

Nvidia might be shouting the loudest in the race to autonomous vehicles right now, but let’s put things in perspective; it doesn’t actually mean anything.