Let’s be clear about something up front: The self-driving Uber vehicle that struck and killed Elaine Herzberg in Tempe, Arizona is a tragedy. The 49-year-old pedestrian did nothing wrong and certainly didn’t deserve to die in March.
With that being said, on average more than 100 people die per day in the United States in an auto accident. 100 people every day.
So while this Uber accident certainly will not be overlooked, it also will not likely be a so-called nail-in-the-coffin for self-driving cars. Over the last few weeks, I’ve had many interactions with casual readers and industry observers who say this could be the end for self-driving cars.
I don’t believe that’s the case — not by a long shot.
Driving is common. Millions of us do it every single day to run errands and go to work. I don’t need to explain to anyone how common a vehicle is in the United States. That said, out of those millions, roughly 100 people per day will not survive.
Think of it like the common cold. Millions of people catch a sniffle or cough during the year. Most people — certainly not 100 people per day — don’t die from it. But imagine if they did? And the big pharma and biotech companies like Celgene Corporation (CELG) , Gilead Sciences, Inc. (GILD) , Johnson & Johnson (JNJ) and Bristol-Myers Squibb Co (BMY) were working on a fix.
Say then, a patient dies in one of those experiments. Not because of the cold, but because of the treatment administered by one of the companies. Would it make sense to shut down the whole program? To say, despite the potential, these treatments are too dangerous to refine, knowing darn well that once refined they will save hundreds of thousands and eventually millions of lives?
That’s the same thing with self-driving car technology! We can’t shutter an entire industry due to one accident, because there are too many lives at stake.
Looking Forward With Self-Driving Cars
In light of the Uber accident, the company suspended its self-driving car testing, as did Nvidia Corporation (NVDA) , a holding in Jim Cramer’s Action Alerts PLUS Charitable Trust Portfolio, on public roads. In the words of Arizona governor Doug Ducey, it was an “unquestionable failure” on Uber’s part. The state has since banned self-driving Uber vehicles for the time being.
Unfortunately, Nvidia’s decision became public right in the midst of CEO Jensen Huang’s keynote speech at the company’s GTC conference near the end of March. The same keynote where Huang was introducing a bunch of new products and technologies — including Nvidia’s new DRIVE Constellation program.
It brought me back to a conversation I held with Danny Shapiro, Nvidia’s senior director of automotive. I met with Shapiro for about a half hour during the Detroit Auto Show earlier this year, before the Uber accident.
“We need to decide, how much safer than a human is acceptable,” he said.
Is it 10 times safer? 100 times? That’s a question we have to answer and it’s a psychological issue we have to battle. We subconsciously accept the risk of driving each time we get in the car. When there is a collision and fatality in a no-foul-play scenario, most parties just accept the accident for what it is — an accident.
But when a machine is the one controlling the wheel, there’s a whole different take. Humans can grasp and accept other human-caused accidents, but are not so forgiving when it comes to an A.I. causing the death.
Say autonomous driving results in 3 deaths per day as opposed to the staggering ~100 we face today. Will that be an acceptable amount, even though 3 people on average are still dying each day? Obviously a zero-fate result would be the end goal, but at what point is the tradeoff worth embracing autonomous driving?
“[We are] working with our customers to now train these neural nets to be robust enough so that we can handle situations much, much more safely than any human could drive,” Shapiro added at the time.
Now, a few months later and with Nvidia’s new Constellation platform, it’s clear the company is delivering on that outlook.
Its new system will enhance programmers’ ability to utilize synthetic scenario testing. In a nutshell — and a bit more in depth of a nutshell can be read here — Nvidia’s new platform can allow testers to draw up all sorts of scenarios for their self-driving car programs. It increases the use of realistic scenarios that the car’s program encounters, but without putting them on the road or endangering anyone.
In light of the Uber accident, more companies will be drawn to these types of technologies, where not only is the efficiency improved but the safety profile improves as well.
I’m not saying there shouldn’t be adjustments, modifications or additional precautions taken. I’m only saying that this technology will lead to safer roads and to turn cold on them now would be a foolish decision.