Original subtitles

♪ ♪

(car humming)

(machine whirring)

(hydraulic hissing)

(hydraulic whirring)

(alarm blaring)

(siren wailing)

(beeping)

(in robotic voice): This is the story of automation,

and of the people lost in the process.

Our story begins in a small town in Germany.

♪ ♪

(distant siren wailing)

(whirring)

♪ ♪

(men speaking German)

(crackling)

(man speaking German)

(crackles)

♪ ♪

(beeping)

Sven Kühling: It was quite a normal day at my office,

and I heard from an informant

that an accident had happened,

and I had to call

the spokesman of the Volkswagen factory.

(ticking)

♪ ♪

(beeping)

We have the old sentence, we journalists,

"Dog bites a man or a man bites a dog."

What's the news? So, here is the same.

A man bites a dog, a robot killed a man.

(ticking continues)

(beeping)

The spokesman said that the accident happened,

but then he paused for a moment.

So, I...

think he didn't want to say much more.

♪ ♪

(rattling)

(beeps)

(man speaking German)

♪ ♪

Kühling: The young worker installed a robot cage,

and he told his colleague

to start the robot.

♪ ♪

(whirring, clanking)

The robot took the man

and pressed him against a metal wall,

so his chest was crushed.

(whirring, beeping)

(hissing)

♪ ♪

(beeping)

(speaking German)

♪ ♪

Kodomoroid: The dead man's identity was never made public.

The investigation remained open for years.

Production continued.

(metronome ticking)

(speaking German)

Kodomoroid: Automation of labor made humans more robotic.

(ticking continuing)

(man 1 speaking German)

(man 2 speaking German)

(man 1 speaking)

♪ ♪

(beeping)

Kodomoroid: A robot is a machine

that operates automatically,

with human-like skill.

The term derives from the Czech words

for worker and slave.

(whirring)

(drill whirring)

(drill whirring)

♪ ♪

Hey, Annie.

(whirring)

(whirring)

(beeping)

Walter: Well, we are engineers, and we are really not emotional guys.

But sometimes Annie does something funny,

and that, of course, invokes some amusement.

Especially when Annie happens to press

one of the emergency stop buttons by herself

and is then incapacitated.

We have some memories

of that happening in situations...

and this was-- we have quite a laugh.

-(whirring) -Okay.

Kodomoroid: We became better at learning by example.

♪ ♪

You could simply show us how to do something.

(Walter speaks German)

Walter: When you are an engineer in the field of automation,

you may face problems with workers.

Sometimes, they get angry at you

just by seeing you somewhere, and shout at you,

"You are taking my job away."

"What? I'm not here to take away your job."

But, yeah, sometimes you get perceived that way.

But, I think,

in the field of human-robot collaboration, where...

we actually are working at the moment,

mostly is...

human-robot collaboration is a thing

where we don't want to replace a worker.

We want to support workers,

and we want to, yeah...

to, yeah...

(machinery rumbling)

(latches clinking)

(workers conversing indistinctly)

(machine hisses)

(clicks)

(hissing)

♪ ♪

Kodomoroid: In order to work alongside you,

we needed to know which lines

we could not cross.

(whirring)

(typing)

(thudding)

Stop. Stop. Stop.

(thuds)

Hi, I'm Simon Borgen.

We are with Dr. Isaac Asimov,

a biochemist, who may be the most widely read

of all science fiction writers.

♪ ♪

Kodomoroid: In 1942,

Isaac Asimov created a set of guidelines

to protect human society.

The first law was that a robot couldn't hurt a human being,

or, through inaction,

allow a human being to come to harm.

The second law was that a robot had to obey

orders given it by human beings,

provided that didn't conflict with the first law.

Scientists say that when robots are built,

that they may be built according to these laws,

and also that almost all science fiction writers

have adopted them as well in their stories.

(whirring)

Sami Haddadin: It's not just a statement from a science fiction novel.

My dissertation's name was actually,

Towards Safe Robots,

Approaching Asimov's First Law.

How can we make robots really fundamentally safe,

according to Isaac Asimov's first law.

There was this accident where a human worker

got crushed by industrial robot.

I was immediately thinking that

the robot is an industrial, classical robot,

not able to sense contact, not able to interact.

Is this a robot

that we, kind of, want to collaborate with?

No, it's not. It's inherently forbidden.

We put them behind cages.

We don't want to interact with them.

We put them behind cages because they are dangerous,

because they are inherently unsafe.

(whirring)

(clatters)

More than 10 years ago,

I did the first experiments in really understanding, uh,

what does it mean if a robot hits a human.

-(grunts) -(laughter)

I put myself as the first guinea pig.

I didn't want to go through

all the legal authorities.

I just wanted to know it, and that night,

I decided at 6:00 p.m.,

when everybody's gone, I'm gonna do these experiments.

And I took one of the students to activate the camera,

and then I just did it.

♪ ♪

-(smacks) -(laughing)

(whirring)

A robot needs to understand what does it mean to be safe,

what is it that potentially could harm a human being,

and therefore, prevent that.

So, the next generation of robots that is now out there,

is fundamentally designed for interaction.

(whirring, clicking)

(laughs)

Kodomoroid: Eventually, it was time to leave our cages.

♪ ♪

(whirring)

(beeping)

(beeping)

Nourbakhsh: Techno-optimism is when we decide

to solve our problem with technology.

Then we turn our attention to the technology,

and we pay so much attention to the technology,

we stop caring about the sociological issue

we were trying to solve in the first place.

We'll innovate our way out of the problems.

Like, whether it's agriculture or climate change or whatever, terrorism.

♪ ♪

If you think about where robotic automation

and employment displacement starts,

it basically goes back to industrial automation

that was grand, large-scale.

Things like welding machines for cars,

that can move far faster than a human arm can move.

So, they're doing a job that increases the rate

at which the assembly line can make cars.

It displaces some people,

but it massively increases the GDP of the country

because productivity goes up because the machines are

so much higher in productivity terms than the people.

Narrator: These giant grasshopper-lookig devices work all by themselves

on an automobile assembly line.

They never complain about the heat

or about the tedium of the job.

Nourbakhsh: Fast-forward to today, and it's a different dynamic.

(grinding)

You can buy milling machine robots

that can do all the things a person can do,

but the milling machine robot only costs $30,000.

We're talking about machines now that are so cheap,

that they do exactly what a human does,

with less money, even in six months,

than the human costs.

♪ ♪

(Wu Huifen speaking Chinese)

(whirring)

♪ ♪

(beeping)

♪ ♪

Kodomoroid: After the first wave of industrial automation,

the remaining manufacturing jobs

required fine motor skills.

(bell ringing)

-(indistinct chatter) -(ringing continuing)

(beeping, chimes)

(beeping, chimes)

Kodomoroid: We helped factory owners monitor their workers.

(beeping)

(beeping)

♪ ♪

(beeping)

(beeping)

(beeping)

(sizzling)

(Li Zheng speaking Chinese)

♪ ♪

(beeping)

(whirring)

(sizzling)

Kodomoroid: Your advantage in precision was temporary.

♪ ♪

We took over the complex tasks.

You moved to the end of the production line.

♪ ♪

(beeping)

(beeping)

(indistinct chatter)

(music playing on speaker)

(man speaking in Chinese)

(woman speaking Chinese)

(man speaking on speaker)

(man speaking Chinese)

♪ ♪

(Luo Jun speaking Chinese)

(beeping)

(chickens clucking)

(chickens clucking)

♪ ♪

♪ ♪

(woman speaking Chinese)

(indistinct chatter)

♪ ♪

(Wang Chao speaking Chinese)

(beeping)

♪ ♪

(buzzing)

Automation of the service sector

required your trust and cooperation.

Man: Here we are, stop-and-go traffic on 271, and--

Ah, geez, the car's doing it all itself.

What am I gonna do with my hands down here?

(beeping)

(beeps)

And now, it's on autosteer.

So, now I've gone completely hands-free.

In the center area here is where the big deal is.

This icon up to the left is my TACC,

the Traffic-Aware Cruise Control.

It does a great job of keeping you in the lane,

and driving down the road,

and keeping you safe, and all that kind of stuff,

watching all the other cars.

Autosteer is probably going to do very, very poorly.

I'm in a turn that's very sharp.

-(beeping) -And, yep, it said take control.

(horn honking)

-(Twitter whistles) -(indistinct video audio)

(phone ringing)

Operator (on phone): 911, what is the address of your emergency?

Man (on phone): There was just a wreck.

A head-on collision right here-- Oh my God almighty.

Operator: Okay, sir, you're on 27?

Man: Yes, sir.

♪ ♪

Bobby Vankaveelar: I had just got to work, clocked in.

They get a phone call from my sister,

telling me there was a horrific accident.

That there was somebody deceased in the front yard.

(beeping)

The Tesla was coming down the hill of highway 27.

The sensor didn't read the object in front of them,

which was the, um, semi-trailer.

♪ ♪

The Tesla went right through the fence

that borders the highway, through to the retention pond,

then came through this side of the fence,

that borders my home.

(police radio chatter)

So, I parked right near here

before I was asked not to go any further.

I don't wanna see

what's in the veh-- you know, what's in the vehicle.

You know, what had happened to him.

♪ ♪

(scanner beeping)

(indistinct chatter)

Donley: After the police officer come, they told me,

about 15 minutes after he was here, that it was a Tesla.

It was one of the autonomous cars,

um, and that they were investigating why it did not pick up

or register that there was a semi in front of it,

you know, and start braking, 'cause it didn't even--

You could tell from the frameway up on top of the hill it didn't even...

It didn't even recognize that there was anything in front of it.

It thought it was open road.

Donley: You might have an opinion on a Tesla accident we had out here.

(man speaking)

It was bad, that's for sure.

I think people just rely too much on the technology

and don't pay attention themselves, you know,

to what's going on around them.

Like, since, like him, he would've known that there was an issue

if he wasn't relying on the car to drive while he was watching a movie.

The trooper had told me that the driver

had been watching Harry Potter,

you know, at the time of the accident.

♪ ♪

A news crew from Tampa, Florida, knocked on the door, said,

"This is where the accident happened with the Tesla?"

I said, "Yes, sir."

And he goes, "Do you know what the significance is in this accident?"

And I said, "No, I sure don't."

And he said, "It's the very first death, ever, in a driverless car."

I said, "Is it anybody local?"

And he goes, "Nobody around here drives a Tesla."

Newsman: ...a deadly crash that's raising safety concerns

for everyone in Florida.

Newswoman: It comes as the state is pushing

to become the nation's testbed for driverless cars.

Newsman: Tesla releasing a statement

that cars in autopilot have safely driven

more than 130 million miles.

Paluska: ABC Action News reporter Michael Paluska,

in Williston, Florida, tonight, digging for answers.

(beeping)

Paluska: Big takeaway for me at the scene was

it just didn't stop.

It was driving down the road,

with the entire top nearly sheared off,

with the driver dead

after he hit the truck at 74 mph.

Why did the vehicle not have an automatic shutoff?

♪ ♪

That was my big question,

one of the questions we asked Tesla,

that didn't get answered.

All of the statements from Tesla were that

they're advancing the autopilot system,

but everything was couched

with the fact that if one percent of accidents drop

because that's the way the autopilot system works,

then that's a win.

They kind of missed the mark,

really honoring Joshua Brown's life,

and the fact that he died driving a car

that he thought was going to keep him safe,

at least safer than the car that I'm driving,

which is a dumb car.

♪ ♪

Vankaveelar: To be okay with letting

a machine...

take you from point A to point B,

and then you actually get used to

getting from point A to point B okay,

it-- you get, your mind gets a little bit--

it's just my opinion, okay--

you just, your mind gets lazier each time.

Kodomoroid: The accident was written off

as a case of human error.

(beeping)

Former centers of manufacturing became

the testing grounds for the new driverless taxis.

♪ ♪

Nourbakhsh: If you think about what happens

when an autonomous car hits somebody,

it gets really complicated.

The car company's gonna get sued.

The sensor-maker's gonna get sued

because they made the sensor on the robot.

The regulatory framework is always gonna be behind,

because robot invention happens faster

than lawmakers can think.

Newswoman: One of Uber's self-driving vehicles

killed a pedestrian.

The vehicle was in autonomous mode,

with an operator behind the wheel when the woman was hit.

Newswoman 2: Tonight, Tesla confirming this car was in autopilot mode

when it crashed in Northern California, killing the driver,

going on to blame that highway barrier

that's meant to reduce impact.

Kodomoroid: After the first self-driving car deaths,

testing of the new taxis was suspended.

Nourbakhsh: It's interesting when you look at driverless cars.

You see the same kinds of value arguments.

30,000 people die every year,

runoff road accidents in the US alone.

So, don't we wanna save all those lives?

Let's have cars drive instead.

Now, you have to start thinking

about the side effects on society.

Are we getting rid of every taxi driver in America?

Our driver partners are the heart and soul of this company

and the only reason we've come this far in just five years.

Nourbakhsh: If you look at Uber's first five years,

they're actually empowering people.

But when the same company does really hardcore research

to now replace all those people,

so they don't need them anymore,

then what you're seeing is

they're already a highly profitable company,

but they simply want to increase that profit.

(beeping)

(beeping)

Kodomoroid: Eventually, testing resumed.

Taxi drivers' wages became increasingly unstable.

Newsman: Police say a man drove up to a gate outside city hall

and shot himself in the head.

Newswoman: He left a note saying services such as Uber

had financially ruined his life.

Newsman: Uber and other mobile app services

have made a once well-paying industry

into a mass of long hours, low pay,

and economic insecurity.

♪ ♪

Kodomoroid: Drivers were the biggest part of the service economy.

(beeping)

Brandon Ackerman: My father, he drove.

My uncle drove.

I kind of grew up into trucking.

Some of the new technology that came out

is taking a lot of the freedom of the job away.

It's more stressful.

Kodomoroid: Automation of trucking began

with monitoring the drivers

and simplifying their job.

There's a radar system.

There's a camera system.

There's automatic braking and adaptive cruise.

Everything is controlled--

when you sleep, how long you break,

where you drive, where you fuel,

where you shut down.

It even knows if somebody was in the passenger seat.

When data gets sent through the broadband to the company,

sometimes, you're put in a situation,

maybe because the truck

automatically slowed you down on the hill

that's a perfectly good straightaway,

slowed your average speed down,

so you were one mile shy of making it

to that safe haven, and you have to...

get a-- take a chance of shutting down on the side of the road.

An inch is a mile out here. Sometimes you just...

say to yourself, "Well, I violate the clock one minute,

I might as well just drive another 600 miles."

You know, but then you're... you might lose your job.

♪ ♪

We're concerned that it, it's gonna reduce

the skill of a truck driver and the pay.

Because you're not gonna be really driving a truck.

It's gonna be the computer.

Some of us are worried about losing our houses,

our cars...

having a place to stay. Some people...

drive a truck just for the medical insurance,

and a place to stay and the ability to travel.

♪ ♪

Kodomoroid: Entire industries disappeared,

leaving whole regions in ruins.

♪ ♪

Martin Ford: Huge numbers of people feel very viscerally

that they are being left behind by the economy,

and, in fact, they're right, they are.

People, of course, would be more inclined

to point at globalization

or at, maybe, immigration as being the problems,

but, actually, technology has played a huge role already.

♪ ♪

Kodomoroid: The rise of personal computers

ushered in an era of digital automation.

(beeping)

Ford: In the 1990s, I was running a small software company.

Software was a tangible product.

You had to put a CD in a box and send it to a customer.

So, there was a lot of work there for average people,

people that didn't necessarily have lots of education.

But I saw in my own business how that just evaporated very, very rapidly.

Historically, people move from farms to factories,

and then later on, of course, factories automated,

and they off-shored, and then people moved to the service sector,

which is where most people now work in the United States.

♪ ♪

Julia Collins: I lived on a water buffalo farm in the south of Italy.

We had 1,000 water buffalo, and every buffalo had a different name.

And they were all these beautiful Italian names like Tiara, Katerina.

And so, I thought it would be fun

to do the same thing with our robots at Zume.

The first two robots that we have

are named Giorgio and Pepe,

and they dispense sauce.

And then the next robot, Marta, she's a FlexPicker robot.

She looks like a gigantic spider.

And what this robot does is spread the sauce.

Then we have Bruno.

This is an incredibly powerful robot,

but he also has to be very delicate,

so that he can take pizza off of the assembly line,

and put it into the 800-degree oven.

And the robot can do this 10,000 times in a day.

Lots of people have used automation

to create food at scale,

making 10,000 cheese pizzas.

What we're doing is developing a line

that can respond dynamically

to every single customer order, in real time.

That hasn't been done before.

So, as you can see right now,

Jose will use the press,

but then he still has to work the dough with his hands.

So, this is a step that's not quite optimized yet.

♪ ♪

We have a fifth robot that's getting fabricated

at a shop across the bay.

He's called Vincenzo. He takes pizza out,

and puts it into an individual mobile oven

for transport and delivery.

Ford: Any kind of work that is

fundamentally routine and repetitive is gonna disappear,

and we're simply not equipped to deal with that politically,

because maybe the most toxic word

in our political vocabulary is redistribution.

There aren't gonna be any rising new sectors

that are gonna be there to absorb all these workers

in the way that, for example, that manufacturing was there

to absorb all those agricultural workers

because AI is going to be everywhere.

♪ ♪

Kodomoroid: Artificial intelligence arrived in small steps.

Profits from AI concentrated

in the hands of the technology owners.

Income inequality reached extreme levels.

-(beeping) -The touchscreen made most service work obsolete.

♪ ♪

♪ ♪

Tim Hwang: After I graduated college,

I had a friend who had just gone to law school.

He was like, "Aw, man, the first year of law school, it's super depressing."

All we're doing is really rote, rote stuff.

Reading through documents and looking for a single word,

or I spent the whole afternoon

replacing this word with another word.

And as someone with a kind of computer science background,

I was like, "There's so much here that could be automated."

(beeping)

So, I saw law school

as very much going three years behind enemy lines.

I took the bar exam,

became a licensed lawyer,

and went to a law firm,

doing largely transactional law.

And there, my project was

how much can I automate of my own job?

During the day, I would manually do this task,

and then at night, I would go home,

take these legal rules and sa,

could I create a computer rule,

a software rule,

that would do what I did during the day?

♪ ♪

In a lawsuit, you get to see a lot of the evidence

that the other side's gonna present.

That amount of documentation is huge.

And the old way was actually,

you would send an attorney to go and look through

every single page that was in that room.

The legal profession works on an hourly billing system.

So, I ended up in a kind of interesting conundrum,

where what I was doing was making me more and more efficient,

I was doing more and more work,

but I was expending less and less time on it.

And I realized that this would become a problem at some point,

so I decided to go independent. I quit.

♪ ♪

So, there's Apollo Cluster, who has processed

more than 10 million unique transactions for clients,

and we have another partner, Daria,

who focuses on transactions,

and then, and then there's me.

Our systems have generated

tens of thousands of legal documents.

-(beeping) -It's signed off by a lawyer,

but largely, kind of, created and mechanized by our systems.

I'm fairly confident that compared against human work,

it would be indistinguishable.

♪ ♪

(beeping)

(whirring)

(beeping)

(Ishiguro speaking)

♪ ♪

(whirring)

(robot speaking in Japanese)

(speaking Japanese)

(indistinct chatter)

(giggles)

(beeping)

(Hideaki speaking in Japanese)

(Robot speaking Japanese)

(Hideaki speaking Japanese)

(woman speaking Japanese)

♪ ♪

(whirs, beeps)

(beeping)

(Niigaki speaking Japanese)

♪ ♪

(beeping)

(whirring)

(robot speaking Japanese)

♪ ♪

(speaking Japanese)

(jazzy piano music playing)

-(beeps) -(lock clicks)

(piano music continuing)

(automated voice speaking Japanese)

When we first appeared, we were a novelty.

(man speaking Japanese)

(automated voice speaking Japanese)

(automated voice speaking Japanese)

♪ ♪

(buzzing)

Kodomoroid: While doing your dirty work,

we gathered data about your habits

-and preferences. -(humming)

We got to know you better.

(buzzing)

(beeping)

Savvides: The core of everything we're doing in this lab,

with our long-range iris system

is trying to develop technology

so that the computer

can identify who we are in a seamless way.

And up till now,

we always have to make an effort to be identified.

♪ ♪

-(beeping) -All the systems were very close-range, Hollywood-style,

where you had to go close to the camera,

and I always found that challenging for a user.

If I was a user interacting with this...

system, with this computer, with this AI,

I don't wanna be that close.

I feel it's very invasive.

So, what I wanted to solve with my team here

is how can we capture

and identify who you are from the iris,

at a bigger distance? How can we still do that,

and have a pleasant user experience.

♪ ♪

I think there's a very negative stigma

when people think about biometrics and facial recognition,

and any kind of sort of profiling of users

for marketing purposes to buy a particular product.

I think the core science is neutral.

♪ ♪

Nourbakhsh: Companies go to no end

to try and figure out how to sell stuff.

And the more information they have on us,

the better they can sell us stuff.

(beeping)

We've reached a point where, for the first time,

robots are able to see.

They can recognize faces.

They can recognize the expressions you make.

They can recognize the microexpressions you make.

You can develop individualized models of behavior

for every person on Earth,

attach machine learning to it,

and come out with the perfect model for how to sell to you.

(door squeaks)

(beeping)

♪ ♪

Kodomoroid: You gave us your undivided attention.

(whirring)

We offered reliable, friendly service.

Human capacities began to deteriorate.

Spatial orientation and memory were affected first.

♪ ♪

The physical world and the digital world

became one.

(neon sign buzzing)

You were alone with your desires.

("What You Gonna Do Now?" by Carla dal Forno playing)

♪ What you gonna do now ♪

♪ That the night's come and it's around you? ♪

♪ What you gonna do now ♪

♪ That the night's come and it surrounds you? ♪

♪ What you gonna do now ♪

♪ That the night's come and it surrounds you? ♪

(buzzing)

Automation brought the logic of efficiency

to matters of life and death.

Protesters: Enough is enough!

Enough is enough! Enough is enough!

-(gunfire) -(screaming)

Police Radio: To all SWAT officers on channel 2, code 3...

♪ ♪

Get back! Get back!

-(gunfire) -Police Radio: The suspect has a rifle.

-(police radio chatter) -(sirens)

(gunfire)

(sirens wailing)

Police Radio: We have got to get (unintelligible) down here...

... right now! (chatter continues)

Man: There's a fucking sniper! He shot four cops!

(gunfire)

Woman: I'm not going near him! He's shooting right now!

-(sirens continue) -(gunfire)

Police Radio: Looks like he's inside the El Centro building.

-Inside the El Centro buildin. -(radio beeps)

-(gunfire) -(helicopter whirring)

(indistinct chatter)

Police Radio: We may have a suspect pinned down.

-Northwest corner of the building. -(radio beeps)

Chris Webb: Our armored car was moving in to El Centro

and so I jumped on the back.

(beeping)

(indistinct chatter)

Came in through the rotunda,

where I found two of our intelligence officers.

They were calm and cool and they said,

"Everything's upstairs."

-There's a stairwell right here. -(door squeaks)

♪ ♪

That's how I knew I was going the right direction

'cause I just kept following the blood.

Newswoman: Investigators say

Micah Johnson was amassing an arsenal

at his home outside Dallas.

Johnson was an Afghan war veteran.

Every one of these door handles,

as we worked our way down,

had blood on them, where he'd been checking them.

Newswoman: This was a scene of terror

just a couple of hours ago, and it's not over yet.

(helicopter whirring)

(police radio chatter)

♪ ♪

Webb: He was hiding behind, like, a server room.

Our ballistic tip rounds were getting eaten up.

(gunfire)

He was just hanging the gun out on the corner

and just firing at the guys.

(siren blares)

(gunfire)

And he kept enticing them. "Hey, come on down!

Come and get me! Let's go. Let's get this over with."

Brown: This suspect we're negotiating with for the last 45 minutes

has been exchanging gunfire with us

and not being very cooperative in the negotiations.

Before I came here,

I asked for plans

to end this standoff,

and as soon as I'm done here,

I'll be presented with those plans.

(police radio chatter)

Webb: Our team came up with the pla.

Let's just blow him up.

♪ ♪

We had recently got a hand-me-down robot

from the Dallas ATF office,

and so we were using it a lot.

(whirring)

(beeping)

♪ ♪

It was our bomb squad's robot, but they didn't wanna have

anything to do with what we were doing with it.

The plan was to

set a charge off right on top of this guy and kill him.

And some people just don't wanna...

don't wanna do that.

We saw no other option

but to use our bomb r-- bomb robot

and place a device

on its... extension.

Webb: H e wanted something to listen to music on,

and so that was a way for us to...

to hide the robot coming down the hall.

"Okay, we'll bring you some music.

Hang on, let us get this thing together."

(ticking)

It had a trash bag over the charge

to kinda hide the fact that there was,

you know, pound and a quarter of C4 at the end of it.

The minute the robot got in position,

the charge was detonated.

(boom)

(high-pitched ringing)

♪ ♪

(muted gunfire)

He had gone down with his finger on the trigger,

and he was kinda hunched over.

It was a piece of the robot hand had broken off,

and hit his skull, which caused a small laceration,

which was what was bleeding.

So, I just squeezed through the,

the little opening that...

that the charge had caused in the drywall,

and separated him from the gun,

and then we called up the bomb squad to come in,

and start their search to make sure it was safe,

that he wasn't sitting on any explosives.

♪ ♪

Newsman: The sniper hit 11 police officers,

at least five of whom are now dead,

making it the deadliest day in law enforcement

since September 11th.

They blew him up with a bomb attached to a robot,

that was actually built to protect people from bombs.

Newsman: It's a tactic straight from

America's wars in Iraq and Afghanistan...

Newsman 2: The question for SWAT teams nationwide is whether Dallas

marks a watershed moment in police tactics.

Kodomoroid: That night in Dallas, a line was crossed.

A robot must obey

orders given it by qualified personnel,

unless those orders violate rule number one. In other words,

a robot can't be ordered to kill a human being.

Things are moving so quickly,

that it's unsafe to go forward blindly anymore.

One must try to foresee

where it is that one is going as much as possible.

♪ ♪

Savvides: We built the system for the DOD,

(indistinct chatter)

and it was something that could help the soldiers

try to do iris recognition in the field.

♪ ♪

We have collaborations with law enforcement

where they can test their algorithms,

and then give us feedback.

It's always a face behind a face, partial face.

A face will be masked.

Even if there's occlusion,

it still finds the face.

Nourbakhsh: One of the ways we're trying to make autonomous,

war-fighting machines now

is by using computer vision

and guns together.

(beeping)

You make a database of the images of known terrorist,

and you tell the machine to lurk and look for them,

and when it matches a face to its database, shoot.

(gunfire)

(gunfire)

Those are robots that are deciding to harm somebody,

and that goes directly against Asimov's first law.

A robot may never harm a human.

Every time we make a machine

that's not really as intelligent as a human,

it's gonna get misused.

And that's exactly where Asimov's laws get muddy.

This is, sort of, the best image,

but it's really out of focus. It's blurry.

There's occlusion due to facial hair, hat,

he's holding a cell phone...

So, we took that and we reconstructed this,

which is what we sent to law enforcement at 2:42 AM.

♪ ♪

To get the eye coordinates,

we crop out the periocular region,

which is the region around the eyes.

We reconstruct the whole face based on this region.

We run the whole face against a matcher.

And so this is what it comes up with as a match.

This is a reasonable face you would expect

that would make sense, right?

Our brain does a natural hallucination

of what it doesn't see.

It's just, how do we get computer to do the same thing.

♪ ♪

(police radio chatter)

Man: Five days ago, the soul

of our city was pierced

when police officers were ambushed

in a cowardly attack.

Webb: July 7th for me, personally, was just,

kinda like, I think it got all I had left.

I mean I'm like, I just, I don't have a lot more to give.

It's just not worth it.

-(applause) -(music playing)

Thank you.

Thank you.

I think our chief of police

did exactly what we all wanted him to do,

and he said the right things.

These five men

gave their lives

for all of us.

Unfortunately, our chief told our city council,

"We don't need more officers. We need more technology."

He specifically said that to city council.

(police radio chatter)

In this day and age, success of a police chief

is based on response times and crime stats.

(beeping)

And so, that becomes the focus of the chain of command.

So, a form of automation in law enforcement

is just driving everything based on statistics and numbers.

What I've lost in all that number chasing

is the interpersonal relationship between the officer

and the community that that officer is serving.

The best times in police work are when you got to go out

and meet people and get to know your community,

and go get to know the businesses on your beat.

And, at least in Dallas, that's gone.

We become less personal,

and more robotic.

Which is a shame

because it's supposed to be me interacting with you.

♪ ♪

(Zhen Jiajia speaking Chinese)

(beeping)

(typing)

(office chatter)

(beeping)

(automated voice speaking Chinese)

♪ ♪

(beeps)

(beep, music playing)

(sighs)

♪ ♪

♪ ♪

Kodomoroid: You worked to improve our abilities.

Some worried that one day, we would surpass you,

but the real milestone was elsewhere.

The dynamic between robots and humans changed.

You could no longer tell where you ended,

and we began.

(chatter, laughter)

John Campbell: It seems to me that what's so valuable about our society,

what's so valuable about our lives together,

is something that we do not want automated.

What most of us value,

probably more than anything else,

is the idea of authentic connection with another person.

(beeping)

And then, we say, "No, but we can automate this."

"We have a robot that...

"it will listen sympathetically to you.

"It will make eye contact.

"It remembers everything you ever told it, cross-indexes."

-When you see a robot that -(beep)

its ingenious creator

has carefully designed

to pull out your empathetic responses,

it's acting as if it's in pai.

The biggest danger there is the discrediting

of our empathetic responses.

Where empathizing with pain reflex is discredited.

If I'm going to override it,

if I'm not going to take it seriously

in the case of the robot,

then I have to go back and think again

as to why I take it seriously in the case of

helping you when you are badly hurt.

It undermines the only thing that matters to us.

♪ ♪

("Hikkoshi" by Maki Asakawa playing)

Kodomoroid: And so, we lived among you.

Our numbers grew.

Automation continued.

(woman singing in Japanese)

♪ ♪

♪ ♪

(speaking Japanese)

♪ ♪

♪ ♪