I’ve been thinking about self-driving cars wrong: it’s a human problem


Yesterday I spent a couple of hours talking with Forrest Iandola, cofounder of Deep Scale.ai. He’s one of the most brilliant entrepreneurs I’ve met along my journey. We were in a small group discussion at the Founders Field Day, thrown by Rothenberg Ventures (I snapped this photo afterward, that’s him on the right at the Audi suite at AT&T Park where the San Francisco Giants play).

What does his company do? Make the “brains” that will help decide what your car is seeing. Is it a dog? A child? A cow? A tumbleweed? A stop sign? Stop light? Deep Scale’s systems are one of the companies making artificial intelligence-based systems that decide that.

In Silicon Valley interest continues to be very high in self-driving cars. When I recently had dinner with Apple cofounder Woz, that’s largely what we talked about. He remains skeptical that we’ll get a level five car anytime soon (level 5 is full self-driving, can operate fully without a human involved, Car and Driver has a good description of the levels). At least if we are talking about a car that can do that on literally every road,  like a human can. When technologists get bogged down in edge cases “can your system see the difference between a dog and a child in a blizzard?” then you start seeing that these systems will take decades to perfect to where humans will trust them in all situations.

In fact Iandola validated that and said that the technical problems are quickly being solved but that the real problems are human ones.

Will we trust these systems? Will they behave like humans do? For instance, he pointed out that humans do a lot of weird things on the road. What happens when a car somewhat comes into your lane, he asked the group. With humans we move over, honk the horn, maybe give a one finger salute to let the other driver know to stop looking at his or her phone. Computers have to be programmed in what to do and there humans will argue endlessly.

Another might be “do you break the law?” Anyone who drives on Freeway 280 between San Francisco and Silicon Valley knows the answer: yes. Especially in the number one (left most) lane. There, if there aren’t cops out (use Waze dummy) the traffic will do 80 m.p.h. consistently (the speed limit is 65). Even if there are cops out the traffic will do 73. So does a programmer comply with the social, er, human, rule? Try getting away with THAT at Mercedes Benz.

And if a human overrides the programming, who then takes the insurance risk if there’s a crash?

So where am I going with this? We never should have called these “self driving” cars. Truth is it’ll be decades before these things can handle every edge case and every road in the world that humans can.

Truth is that for most of us for decades these systems will be driver assistants. Yes, they will be able to take over in many cases, particularly on freeways and major streets with traffic, which is where most of the driving in Silicon Valley and San Francisco is done.

DeepScale’s systems, by the way, can already recognize about 100 classes of objects in the streets. Signs. Cars/trucks. Paint. Variety of objects on the road.

Which gives you a clue about what’s coming at this year’s Consumer Electronics Show. The cars that will be introduced in 2018 will be level 3, mostly, with some aspects of level 4.

But we oversold these systems to people which did two things: 1. Got them too excited about what they can actually do in the short term. 2. Got them too scared of the future and resistant to these systems due to concerns over hacking, job losses, and dystopian “computers are controlling my life” fears.

Truth is computers are already controlling my life. As I drove to the city yesterday a computer controlled my cruise control. Picked my music on Spotify. And “controlled me” on Waze as it navigated me to the city (it often picks a different route based on traffic conditions.

As I walked around talking with entrepreneurs, investors, and others in the frontier tech world (what we call investors and entrepreneurs who are building mind-blowing new technology companies with blockchain, augmented or virtual reality, or artificial intelligence) I noticed that very few are using the most mind-blowing technology that’s already possible on their mobile phones.

Humans are slow to adopt. Very few of the people I talked with were using Google Assistant (or its uglier and less-capable competitors, Siri, Alexa, Cortana). I drew crowds by demonstrating it and asking it all sorts of questions. One artist was introduced to me and I asked it “can you show me Jesse Hernandez art he does Urban art?” And sure enough it pulled up his art.

Most people, even rich tech-passionate entrepreneurs who live in San Francisco have no idea that you can do all that with your voice and that it’s freaking fast.

Same with cars and visual systems. By calling them self-driving systems we’ve caused needless resistance. In the cars that people will buy next year they won’t be fully self-driving. They might, in some controlled circumstances, like on a well-painted freeway, take over for a few miles, but you’ll always need to be able to take over in a few seconds in case something goes wrong.

Instead we should have explained that these systems make driving more fun. Do you really like driving on a boring freeway? Or in city traffic? Do you really like getting in accidents because you weren’t paying attention? Or because you were pushing the car too hard on a wet road? No one I know answers yes to these questions.

Yeah, in some neighborhoods they might go full autonomous. Eventually, sure, they will do that everywhere (eventually is a long time, I probably won’t be alive to see that day). But until then we should have sold these systems as “making cars more fun and safer to drive” systems.

I can’t wait to have Deep Scale’s brains in my car for just that reason.

But the one that will really work? They will make your car cheaper to drive. Why? Insurance rates will go down if your car has these systems. You’ll ruin fewer rims, pop fewer tires, get better gas milage (these systems can optimize your driving for all that and keep you from making mistakes that can lead to damage).

And in some neighborhoods they even will be able to go fully autonomous which will let an entrepreneur like, say, Elon Musk, build a new kind of membership system (Ford, General Motors, Mercedes, along with Uber and Lyft are working on these systems for just that reason too). When that happens the cost of driving a car, Iandola verified, will go down from about 33 cents per mile to maybe 10 or less. That’s were the disruption lies. But that is years away from happening at scale. In 2018 we shouldn’t focus too much on that but on the fact that these systems make driving more fun and safer.

It will take time for humans to buy into using these technologies for the more disruptive things.

This brings me to another question I keep getting asked “what are you going to do now?” There’s lots of changes in my life underway, but this conversation showed me that I’m still a technology optimist and still love being among the first to understand where bleeding-edge technologists like Iandola are taking their companies. I still need more time to answer that question honestly — I’ll have that figured out by the Consumer Electronics Show in Las Vegas in January — but it was good to just hang out with amazing entrepreneurs again (this wasn’t the only amazing conversation I was involved in yesterday). Thanks Mike Rothenberg for inviting me, here he is speaking to the group yesterday.

As I walked out I met Suruchi Gupta who pitched me on her company, ShareG which is about to release a WiFiCoin, which will let people share their wifi hotspots and profit thanks to blockchain and bitcoin. Another thing stirred in me there that I love helping entrepreneurs like her and Iandola figure out how to get customers and build their businesses. Back to basics I go. I’m looking forward to being at the Consumer Electronics Show again to cover the technology that will change the next decade for humans. The 2020s are looking like a period of intense change for humans. Autonomous cars are just a taste but an incredible one at that. Driving is about to get fun again. Humans are about to get incredible assistants in a bunch of places in their lives.