The State of Autonomous Driving | udelv CEO Daniel Laury

I want to spend a couple of minutes on what is a deep neural network. This part that I was mentioning about algorithm and data, if that works. It doesn't. So what do we call deep neural networks? Deep neural networks enable to— enables the computer through this process of what's called artificial intelligence and machine learning to go from an image that the camera will see and to be able to deduct that this is a vehicle. Sometimes it's as good as being able to detecting this is a regular vehicle or it's an emergency vehicle. And the way this is done is no longer a rule-based computing because there are so many different types of vehicles and so many different angles we could look at them in the world that it'd be— it meant— it would mean millions of pictures for the computer to crunch every millisecond. So it's too much. It's the way the neural network will work is that it will go through iterations. And start analyzing some pixels of the picture. And then go from one iteration to the next one, to the next one, to the next one, to finally come to a probability, stochastic probability, a percentage of saying there's 98% chances this is a vehicle. Below a certain probability, the computer is gonna start recalculating from a different angle, a different image, and reprocess the whole calculation until it gets to a very, very high probability of knowing it's what it is. And this is being done through humongous datasets, very, very large databases of data where basically We teach the computer, we feed the computer with pictures of anything you can think of that you could see in the streets. Automobiles, dogs, plants, ships. You know, it happens less on the street. But, you know, and then we train the computer to understand, you know, that if you see that kind of pixel, if you detect that kind of pixel in an image, eventually this is going to be this object or that object. And that's how perception in the world of autonomous driving works. So what's the state of autonomous driving today? Well, in the last 3 or 4 or 5 years, you know, the industry has grown considerably. You now have hundreds of companies working on autonomous driving. Some building the whole stack like Waymo. and Google, to some others focusing on hardware or software on some parts of artificial intelligence, development tools, sensors, connectivity, mapping, et cetera, et cetera. And this industry is growing. And I predict that it's going to carry the world, the United States and the world, for probably the next 3 decades. There's such a huge ecosystem around this that is in the process of being built as we speak. Now the difficulty of autonomous driving is the following one. The premise behind autonomous driving is that it's— the number one premise is that it's going to save lives. It's going to be safer than cars. Now there's something that sometimes people don't understand at this point is that The premise behind autonomous driving is not to reduce the number of fatalities on the road down to zero. No one has ever thought or said that this is ever going to happen. There will be fatalities. But the premise is that it's going to cut the number of fatalities dramatically. There are currently 40,000 fatalities every year on US roads and nearly 1.4 3 million every year in the world. It's a gigantic number. I mean, compared to that, pretty much everything else is nothing. But very few people talk about it. So if autonomous driving, if computers were able to save, let's say, half of them, half of those fatalities on the road, that would save 20,000 lives in the United States and it would save 650,000 lives lives around the world. It's huge. But at the same time, this is done over tens of billions of miles being driven. In the United States, there is one fatality over 86 million miles driven. And to put things into perspective, Google, which has been at it for 10 years, has currently driven 7 million miles and is by far the most advanced company in autonomous driving for ridesharing. Now it's accelerating and now they're at about a million miles a month, but a million miles a month is 12 million in a year and I said that there's one fatality in 86 million miles. So the industry has to make— That's a lot of miles. To continue to be making a lot of progress to be able to say it's at least as safe as a human driver. And that's what's going on right now. There's a race in the world to drive more miles, which is the reason why Google, for example, or Waymo announced recently that they're going to be purchasing 62,000 Chrysler Pacifica and 20,000 Jaguars. This is Absolutely humongous. You're talking about $20 to $25 billion that Waymo is going to be spending in the next 3 years. But that's to put a lot of vehicles on the road, collect the data, be able to face the corner cases, and be able to rack up more miles to become as safe or more safe than a human driver. So that's the state of the industry. right now. How can autonomous driving influence new services? There are so many ways that you could think of that autonomous driving is going to change our lives. First of all, you guys might have taken a driver's license. Maybe not all of you. But my kids that are— my younger kids that are 11 and 12 may never have to to take a driver's license in their life. It's going to change our lives and our future considerably. Intel actually predicts that this industry is going to be a $7 trillion industry. Mobility as a service will disrupt pretty much anything that we've seen before— ownership patterns of vehicles, Which is, by the way, the least used asset in the world. 90% of the time that you have a consumer car, it's idle. And it's going to change services in many, many different ways. We can imagine pods for fast casual driving that are going to be driving in the street and getting closer to you as opposed to you going to a restaurant. Onboard beauty salons. Mobile vending machines, mobile healthcare clinics. The mobility as a service business is going to explode with that kind of technology. We probably will see some significant changes in our urban environment. Do we need parking lots? Do we need those parking lots to be placed where they are right now? Is it going to save up a lot of space for maybe green spaces? How cities are going to be redesigned for that kind of industry. We can imagine location-based advertising in the vehicles. The inside of a passenger car is going to be changed completely. You're going to be watching movies or doing things while you're riding a vehicle. And so there's a lot of changes. First big one we think is going to be delivery. And the reason why delivery is going to be probably one of the first applications of autonomous driving is because it doesn't— we don't carry passengers. So if you don't carry passengers, you're a lot safer. At least you don't risk the life of the people inside the car because there's no one inside the car. And you can always think about protecting the environment. There are some other reasons that the delivery vehicle can go slower than a passenger car. And a delivery vehicle can have some route optimization patterns that would not be acceptable for passengers, but that may be acceptable for a delivery vehicle, like going right, right, right, right instead of making an unsignalized or unprotected left turn.