NVIDIA and the Big Unaddressed Problem with Self-Driving Cars

This week, NVIDIA had another design win with its self-driving cars, this time with XPENG’s new G6 coup (they are already in the G9 EV). I’ve been watching reviews of some of China’s latest electric cars. While initially many were of such poor quality they’d never sell anyplace else, some of the latest cars are arguably better than Teslas. While they may lag in performance, they seem to provide a greater luxury experience and more fun at a far lower price than a Tesla, and Tesla is the world’s gold standard when it comes to electric vehicles (or it was). And with NVIDIA’s technology, this is how self-driving should work.

But when it comes to self-driving, Tesla has really hurt the market by presenting its Level 2+ system (basically enhanced cruise control) as self-driving and naming it Autopilot, which resulted in an impressive number of deaths and scared consumers away from the technology. Consumer Reports was not amused.

But Tesla isn’t the only one scaring consumers. The self-driving taxis in places like San Francisco continue to act up publicly and scare people away from the technology just a few years (almost months) before it will be available in a workable form.  

Let’s talk about self-driving, why you should want it, and what still needs to be done.  

NVIDIA Orin and self-driving benefits

NVIDIA Orin is currently the best technology for self-driving cars. What makes it the best is not only its hardware that benefits from NVIDIA’s nearly two-decade focus on AI which has catapulted it to one of the most valuable technology companies in the world, but also NVIDIA’s Drive Sim powered by Omniverse. This simulation platform allows it to train its driving AI in hours to a level that would be equivalent to putting cars on the road for years because the training happens at machine speeds that are significantly faster than real-time.     

Things I’ve gotten hooked on on YouTube are the bad driver videos. I try to see the problem that causes the accident with enough time to avoid it. Watching those videos likely saved my life when, the other day, I was entering an intersection on a green light only to have a semi run the red light going around 60. Regardless of how heavy my electric Jaguar is, the semi would have crushed me if I hadn’t developed a habit of looking for this kind of mistake and being ready to brake. 

Why you should want self-driving

Often, the threat on the video comes from a direction where I either can’t see (back, or blind-side) or where I’m not looking. Even parking lot accidents often occur because you are not looking at where the threat is coming from: pedestrian, a car backing out the same time you are, a car driving too fast in a parking lot, or just being distracted at the wrong time. One time I started backing up while looking at the car’s radio expecting it to be a reverse camera because that is what I’d been doing for years (luckily, I didn’t hit anyone).  

It is in situations like these that self-driving technology could literally save your life. These systems use a variety of sensors to look all around the car and can use vehicle-to-vehicle technology to see things outside of the range of these various sensors.  And if you want to drive yourself, you can as the car will drop down into something called Guardian Angel mode where the AI will take control of the vehicle if it sees a potential accident you haven’t seen.  

This capability will save countless lives. People with kids and/or pets on board, those too addicted to social media or texting, or those that are impaired are all catastrophic accidents waiting to happen. Self-driving technology should be able to mitigate most of that exposure.  

One little distraction and a fun Sunday drive can turn into a catastrophic event where people are permanently crippled or killed. I would personally pay a lot to never have to deal with such an event.  

Wrapping up: What still needs to be done.

Self-driving cars that aren’t self-driving will emerge as the biggest threats. AIs can talk to each other, they don’t drive impaired, and they can help each other avoid accidents. But humans can get road rage when they think they’ve been disrespected, or drive under the influence, and often drive distracted. No matter how good an AI is, it can’t overcome physics. If your car stops to avoid an accident, there is a good chance the human-driven car behind you will not stop in time and you’ll still become a statistic. 

So, what still needs to be done is to begin getting people excited about this technology so that it quickly propagates to all the cars around us as well as our own. The life this technology saves may very well be ours. We need to build demand for self-driving cars because if folks don’t want it, we will remain at risk of losing what is most valuable to us: the lives of our loved ones. I don’t know about you, but I’m rather attached to my life and would rather not lose it, either.