But what would their record be like when everything works as it should?

Here we go, on our first ride in autonomous vehicle, it's just me in the car and the robot.

Everything we do from here on out is going to be the autonomous vehicle taking full control.

Oh, while we were driving across this campus just now, a child ran in front of the car and has promised the autonomous vehicle did its job.

It stopped 1.25 million.

That's the number of people who die in auto accidents every year.

Worldwide.

In the United States, more than 40,000 people were killed in accidents in 2016 6% more than the previous year.

It's much worse than all terrorist attacks, much worse than all playing pressures, and we never really talked about it.

I'm on a mission here.

When I was 18 years old, I lost my best friend in a traffic accident due to a split second bad decision, and I felt it was just such a waste, and as the technologies I felt I gotta fix this.

I actually believe with only self driving cars, there would be no accidents.

Look, cast on text.

They don't distracted drunk.

They pay attention all the time.

There was so much better than people.

That's the goal.

But how do we get there?

For answers?

I went toe Waymo, the self driving arm of Google's parent company in Silicon Valley, where engineers are building technology that has clocked more than four million miles of autonomous driving.

Jamie Widow is the lead systems engineer who came toe Waymo after putting a rover on Mars for NASA.

She also co wrote Way Most Playbook on Autonomous Driving Safety.

We have layers of safety around the vehicle.

We have radars that can see the speed of things, so we know if they're moving or static objects.

We have lasers, lighters that paint a three D image of the world.

They can see a football helmet, two football fields away and they can see 360 degrees around the car.

And all of that allows us to see things and see them very far away and see them at high resolution to make all of those decisions and understanding about our world so you can see the things playing out in front of you and you can slow down and respond safely.

Before I left, I had to ask, What's more difficult getting driverless cars here on Earth or putting a rover on Mars, its self driving cars by far.

So what are we going to know that these cars are safe enough to drive?

I asked a lawyer who's an expert on that very question.

This is one of the really tough questions, not just how safe is safe enough.

But how do we know that the vehicle is that safe?

How do we know?

How will we know?

We don't and we're not sure.

I think that there will be a lot of new issues that automation creates, but ultimately we are ill served by today's transportation system.

We talk about the car traditionally as being this symbol of freedom.

You're not free if you're dead.

This is the part that really freaks people out.

If my car is the one doing the driving, how can I trust it not to kill me or someone I love in a bad situation?

Will the car choose one person over the other.

It's a question philosophers call the trolley problem.

Picture this.

A trolley car is heading down a track toward a family of five.

That air stuck.

You can save them by pulling a lever and changing the course of the train, But there's one man stuck on the other track.

Is it more ethical toe?

Let the family die or sacrifice the one man so they could be saved.

What would a car driven by a computer do?

It's not that the trolley problem is impossible to come up.

It's not, but an error has been made if that were ever to occur.

If you're ever in a situation where you're faced with the trolley problem, the car has made a much more serious error.

It should have been more conservative so that it was never faced with that situation.

In other words, we may be thinking about this all wrong.

If you think about a computer playing chess, it's moving many, many moves ahead more moves than you possibly could.

And analyzing the computer is able to process this environment and anticipate much better than you could and operate with much better accuracy than you can were able to track objects all around the car and detect them with superhuman levels of perception.

This, of course, all assumes that the car is in working order.

But these cars are computers, after all, and any computer can be hacked.