Autonomous Vehicles Q&A | udelv CEO Daniel Laury

Correct me if I'm wrong, but I think Teslas are only able to do like the self-driving if there's like lines on the road, right?

Yeah.

So is yours going to be able to actually self-drive on like a country road that's not even paved and doesn't have like those lines?

Yeah, Tesla is a Level 2 vehicle. They do intelligent lane assist.

It's called—

it's driver assistance basically, essentially on highways. We are level 4 of autonomy, meaning we can stop, we can recognize traffic lights, we can recognize stop signs, we can intelligently drive roads, pretty much any type of roads, although the ones that are not marked are more complicated. So, but the machine is going to be trained on those roads eventually, and then we'll be able to do that. So there's a big difference between Is that because of LiDAR? In particular because we're using some LiDARs. Tesla so far has not been using LiDARs.

Hi. So first of all, I think it's pretty amazing. But from what I understood, the laws in the US is that you need a driver in the vehicle, right? So is this basically the only thing that's stopping you from innovating the self-driving car? Or what is the next step? What's the next step?

So this is perhaps the first time in the industry of the world, or in the history of the world, or the history in the industry where in the United States the legislator is moving faster than the technology.

And it's pretty incredible.

And I really respect a lot the legislators for that. In Arizona, It's totally open for business. You can today go to Arizona and put a vehicle on the street that has no steering wheel and no pedals.

It's allowed.

California just passed a bill on April 2nd, so 2 months ago, allowing companies to apply for a permit to remove the safety driver already. So if the technology was ready, We could have— we could remove the safety driver today in California.

That's pretty cool. How are you going to make sure that it's really the guy who should get the delivery?

We have— there's an app and the— when the merchant loads up the vehicle, the name and the address of the customer is linked to a compartment And so the computer does the reverse when it comes to a customer location and the customer will only see the compartment, will only be able to open the compartment where his or her goods are just by the press of a button.

You have it.

I'll come back.

So how do you think the next 30 years for the economy and the technology is going to be? I read that China by 2025 is going to be completely self-sufficient in manufacturing the technology themselves. So how do you intend to compete with companies that are based in China and only the Chinese are allowed to do it in China, like the companies coming out of there, because they'll be completely self-sufficient to do it?

It's a great question. It's a race. There's no doubt about that. And it's going to be a— a tough race, which is again why it's so important for the US legislature to allow this technology to happen and so to enable us to compete, because the market here is smaller than the Chinese market. Now what I think is that at the same time that the Chinese don't let foreign companies operate on their territory, They also have a market that's so big in their own country that they may not be as interested into going in other countries as quickly as they do in their country, which lets us time to do United States and Europe. So hopefully that's the way it's going to happen. But we got to go fast.

Will you be demonstrating the vehicle And can I take a ride around in it?

No, you won't be able to take a ride, but we'll show you the cargo space, how it works.

Who are your current customers?

We have 7 customers operating right now and 2 that have signed with us and we're going to start operating. We started with a grocery store in Silicon Valley called Draeger's. Market, which is down the street here. And that was our first customer. And so we're doing all the deliveries for Draeger's within a certain radius of the store. Then we added a couple of restaurants. Then we added a florist. Then we added an auto parts distributor, which is a great customer because they do the same routes every day to a series of dealerships. And we have a fruit delivery company. So those are local businesses, and we're at the same time conducting discussions with some of the largest retailers in the world, and some of them we'll be announcing soon in other locations in the United States.

So you mentioned using CNNs, convolutional neural networks, and my company uses them as well. And while we do like them, we sometimes face problems on detection, and especially When we try to transfer the information, the transfer learning, do you ever fear that something might be mixed up?

Yeah.

And I don't know, the brakes don't work because of it. Does that ever bother you?

I'm not too worried about a mechanical problem, like the brakes don't work. Well, in general, Vehicles can brake. So we're not— we're actuating the brakes with a motor, but in general, the brakes will work. What might happen— and it's actually been the case in the Uber crash that you might have heard of in Arizona— is it looks like it was a perception issue. A second and a half before the crash actually happened— The driver was looking at the wrong thing. The computer classified the woman and the bicycle as a non-object. Because the difficulty of those perception issues or the classification is that you don't want to have false negatives or false positives. So the computer has to make a quasi-instant decision as to what it actually sees. So a second and a half, it was classified as a non-object. A second before, so half a second later, they— it recognized something but wasn't able to classify it. And then a half second before the crash, it finally detected that it was an object, but the computer made the decision that it didn't have enough time to brake.

Because it's a passenger car and it might be rear-ended and create a bigger accident.

So it's this famous conundrum which is complicated to solve for passenger cars. Do I kill the pedestrian or do I kill the passengers in the vehicle or do I kill people behind me? And that's a very complicated decision, which is one of the major reasons why I was saying I think delivery is probably going to be one of the first bigger applications because at least we don't have to make that decision. We always protect the pedestrian.

What is the current—

you mentioned electric, like refueled, but it seemed like you also mentioned gas stations.

What's the current—

Yeah, that's a whole different debate. And we're not here to solve it, so we're going with the industry.

When I started this business, and this is why you have this vehicle, is we're in California.

It'd be suicidal not to go with an electric vehicle. It turns out that electric vehicles are not very well suited today with the existing battery technology and the energy density of batteries for first utility vehicles and second autonomous driving utility vehicles. First of all, utility vehicles because the vehicles are supposed to be on the road 80% of the time. Whereas for a passenger car, it's idle 90% of the time, so you have time to recharge.

But for a vehicle that has to be on the road for 80% of the time, you don't have time to recharge.

Second of all, an electric vehicle is still much, much more expensive than a gas vehicle. And stores, retailers, they care about money. They care a lot less about anything else. Third is that when you add to this the computing equipment and the need to operate the sensors and to operate the steering wheel and the pedals, you draw a lot more power than actually the power needed for the propulsion of the vehicle itself. So you end up having the need for a battery that's twice as big, or maybe not twice, but like 1.5 times as big. as the battery that would be needed to propel— for the propulsion of the vehicle, which in turn creates more weight, so then you need a heavier battery for propulsion. So it's a very complicated vicious circle, if you will, that the industry hasn't resolved yet, unfortunately.

I saw an article in Wired magazine He was mentioning Nuro as one of your competitors, I guess.

Yeah. So how is the competition going? So Nuro is a very different vehicle. They got a lot of press attention because they raised a lot of money and they're 2 ex-Googlers.

But it's a very different vehicle.

Nuro started from a sidewalk, being a sidewalk robot like you might have heard of, Starship Technologies. There's at least 20 companies in the world today doing sidewalk robots. And then realized that sidewalks may not be the best idea because you got pedestrians on sidewalk and children and everything. And so they said, how we're going to transform our sidewalk robot to becoming— to going on public roads? And so they created something that's a little bit bigger. It's kind of a hybrid between a sidewalk robot and an actual vehicle that is certified to go on public roads. And it's cool, but it's going very slow. It can carry only 2 customers at a time at best. And also, I'm not sure it's actually certified by the Department of Transportation as a as an actual category of vehicle. Because you might not see it. It's smaller. So if you have a Class 8 truck going closer to this thing, it can roll over it without even seeing it. So I don't know how the DOT and NHTSA is going to classify this kind of vehicle. I don't know about the future of that.

What's the future of the electric vehicle?