so quickly compared to a hundred years ago. But, the cost of that, in the US, 33,000 people

are killed every year. To put that in perspective, it's the leading cause of death for people

between the ages of 4 and 34. It's more dangerous than cancer, it's more dangerous than gun

violence. When you look worldwide, the number's even more scary. It's 1.2 million people killed

every year. 90 plus percent of that is human error. And so if we can bring in technology

that's always paying attention, that can see what's going on around it, that never gets

distracted, this is a huge opportunity. This is one of these 10x kind of opportunities

to save lives and make the world a better place.

People have been thinking about self-driving cars basically since we've had cars. And then

along came the DARPA Grand Challenges in the early 2000s. I and a number of folks on the

team had taken part in those, and they were really exciting. They were kind of like the

Woodstock for Robotics. And then in 2008, Sebastian Thrun, who was at Google, started

talk with some of the leadership here and started to think about the fact that self-driving

cars is really a computer science problem. And it's this huge computer science problem

that will have a big impact on the world, and thus it's exactly the kind of problem

that Google likes to solve. And when we kicked off the project, the question was, is this

even possible? Could you actually have cars that drive themselves out on the road? And

so we set for ourselves two audacious goals. One was to drive 100,000 miles on public roads.

To put that in perspective, this is an order of magnitude more than anyone had ever driven

before. And then the other was to drive 1,000 miles of really interesting places.

After about a year and a half, we were actually able to complete that challenge, and that

told us we actually had technology that could work. Today, one of the things we're most

focused on is making the car drive naturally. We want the car to be on the road, we want

it to have all the aspects of the best human drivers, we want it to be courteous, we want

it to be defensive, but we want it to drive in a way that you can expect how it will behave

on the road because that makes you safe around it as another driver.

If you go back to the Darpa Challenges, the vehicles driving round really were robots.

They would accelerate hard, they would break hard. With our vehicles today, they have to

be good for our occupants, they have to be nice and smooth and safe, and that has to

be very naturalistic. As a human, it's really easy to get uncomfortable. You've probably

experienced this when somebody else drives for you, and you sit in the passenger seat

and they hit the brakes late or they hit the brakes too early, and you get a little tense

about that. The other is from the outside of the car, we want it to feel natural for

the other drivers because when it's natural it's safe. And so our vehicles move very smoothly

on the road, they pay attention, and if anything they're more courteous and more defensive

drivers than normal drivers.

When self-driving cars are a reality, it's going to be amazing. Imagine never losing

someone to a traffic accident again. Imagine a world where you get in your car, it takes

you where you wanna go, and then you get out. And you don't have to search for parking,

you just leave it, and it goes off and helps someone else get where they're going. Imagine

cities where parking garages aren't there, where that land has been turned into into

homes or into parks, it's just gonna be amazing. It's gonna be an exciting place.