Full self-driving capability is just three years away, according to one Silicon Valley expert.
“We’ve probably spent $100 billion in the past four or five years developing technology for fully autonomous go-anywhere cars,” says Blair LaCorte, president of AEye. “That … can now be used in more intelligent ways, in more intelligent business models.”
Quite clearly, not a self-driving car
Photo by Samuele Errico Piccarini on Unsplash
AEye develops advanced vision hardware, software, and algorithms that, the company says, act as the eyes and visual cortex of autonomous vehicles.
I spent 30 minutes with LaCorte talking about autonomous cars, AEye’s technology, and how he sees a path to self-driving cars:
(You can also get a full transcript of our conversation, or subscribe to the Tech First Draft podcast.)
The most critical question, of course, is: how you define an autonomous car?
“We actually have autonomy today,” LaCorte says. “For instance, in level two anti-braking systems, your car decides whether it sees something, and it has the capability because you’ve given it to them, to autonomously decide whether to stop or not. That truly is autonomy, it’s just closed loop autonomy, only doing one type of task.”
Of course, when we talk about self-driving cars, we generally mean open-loop autonomy: a car with the ability to go from just about any A to just about any B, without a human being required to intervene.
Developing and delivering that depends, LaCorte says, on working business models. On the consumer side, with Tesla, that’s the company’s Full Self Driving capability:
“If you actually took out Autopilot out of Tesla they would have no profitability, because their attach rate is somewhere between 30% and 40% for $6000 to $7,000,” LaCorte says. “They were able to help a consumer decide to pay $6000 or $7,000 more for a third of their cars.”
The opportunity here, which Elon Musk has referred to as “robotaxis,” is that the economics of the automotive industry have been turned upside down over the past few generations.
“We have a very expensive, durable product that people only use 4% to 6% of the time,” LaCorte says. “If you take a look at the auto industry from 30 years ago, they sold you a car every three years, the margin was $12,000, they got a 10% kickback from the channel … [but now] the average car company makes $3,000 per car, sells you a car every seven years, and with an EV will be 10 years.”
In other words, GM, Ford, and Toyota aren’t making as much money as they used to, and they’re not going to make it up on volume. Where they can make it up is with services.
Those services include driving you places, but also transportation as a service. If you knew you could have a car whenever you wanted — in other words, more reliable than Uber and Lyft — you’d probably be okay with your car driving other people while you work or sleep. And you might even prefer a model where it’s not your car at all.
Another big opportunity that LaCorte sees, however, is on the B2B side, with companies like UPS and Fedex, who own their own trucks.
“The second level is there will be people who already own an asset today,” LaCorte says. “Who are those people? The bus companies, the people who have campus shuttles or who have construction vehicles … if you can prove that autonomy will make them more money or make them safer and reduce their insurance costs or their downtime, they will deploy.”
In spite of the fact that global investors have poured over $100 billion into self-driving capability, it’s not quite here yet.
A friend tells me his Tesla Model 3 has driven him tens of thousands of miles, but he still needs to pay attention. (And its autopilot may violate state laws.)
And an autonomous semi-truck drove 3,000 miles from California to Pennsylvania to deliver butter in under three days … but it needed a safety driver who was ready to take the wheel, and only drove “mostly” autonomously.
The problem is not sensors, LaCorte says.
The problem is aggregating the data.
“What we’ve tried to do is jump to the end,” LaCorte says, talking about AEye’s self-driving technology. “Car companies don’t want cheaper individual sensors, they want a system that takes whatever sensors they have and allows it to perceive in an intelligent fashion … every system today that from radar, to a camera, to the LIDAR systems … are passive detect systems. They only bring data one way and they bring as much data as they possibly can. The human equivalent in biomimicry is autism.”
In other words, data coming in is good.
But understanding it, and knowing what to pay attention to, is just as important. The reason Uber’s self-driving car killed a man is not that it missed seeing him. It just didn’t decide to pay enough attention and do something about it.
AEye’s technology builds on military-spec technology to not only see objects but also make predictions about what they’ll do and where they’ll go: searching, detecting, acquiring and making decisions, all in the sensor.
“What we’re really showing is once you acquire a target, you can actually predict intent and therefore you can do what a human does: be careful, someone’s going to step off a curb” LaCorte told me. “Be careful, the trunk is open, you’re at a supermarket, someone’s probably loading the trunk. And that will be the big jump again, that will help people feel much, much more comfortable that their cars can be smarter.”
That kind of technology, LaCorte says will make self-driving safer.
And, worth the risk, seeing that traffic accidents cause over a million deaths per year, globally.
Which brings us back to the three years:
“I’m not saying technology is perfect, but all those components for the next three years actually exist, LaCorte says. “The technology doesn’t have to make major jumps. And that’s why I probably am contrarian. If I redefine autonomy, what I’ll tell you is there will be money made on autonomy within three years.”
Unless you’re Tesla, of course.
Tesla has been making money on undelivered self-driving technology for many years already.