[MUSIC PLAYING]

Ethics applies in any context in which you introduce robots.

And it pervades really everything we do.

Robots are entering into a lot more areas of our lives

than ever before and taking on different social roles.

I love robots.

I want to build robots.

But I think we can't be naive about the possible harms

and repercussions that we don't even expect.

Because that will affect real people's lives.

[MUSIC PLAYING]

I think we have a very powerfully ambivalent

relationship to robotics.

And what I mean by that is, on the one hand

we're drawn to them very strongly.

They're very compelling and captivating as objects.

But they also can be very scary in many cases.

[GUNSHOTS]

There's a host of concerns about robotic systems.

Robots really represent the ability

to have roaming, roving cameras that are capturing

all kinds of information.

All of the outdoor space, public space,

places that were not observed, are going to be observed.

And that data is all going to be captured and recorded.

And we really don't have any kind of laws and regulations

about how that can be utilized.

The other thing that makes these very different

is the fact that they move around.

These are physically engaged with the world.

And that means they can do a lot of harm in the physical world.

And Amazon is talking about these quadrotors or octorotors

that are going to be delivering packages to your front door.

My first concern is, well, that's eight lawn mower

blades coming down for a landing in front of your house,

especially if it's carrying a box of books.

Then you have a whole host of legal issues

around who's responsible when these things do some harm.

So there's the person who maybe owns the robot.

There's the person who programs it or maybe

tells it what to do.

The idea of who really shares in these kinds of responsibilities

becomes very critical, especially when

we think about technological innovation.

Because a company that manufactures a robot,

if they hold all of the liability,

then that's a really big impediment

to them bringing to market these different technologies.

But if we shift the law so that they don't have any liability

or we greatly limit the kind of liability that they can have,

we could wind up with a lot of really dangerous robots.

So we really have to kind of find a balance between these.

[MUSIC PLAYING]

Asimov's three laws of robotics are pretty straightforward.

A robot should obey a human being.

A robot shouldn't do any harm.

And a robot should then engage in effect self-preservation.

But if you actually read Asimov's stories,

in nearly every story, something goes awry largely because

of these laws themselves.

So what happens if two humans give orders to the robots

and they're conflicting orders?

What should the robot do then?

He illustrated how just a simple rule-base morality does not

work.

[CRASHING NOISE]

Robots are going to encounter some hair-raising scenarios.

With self-driving cars on a narrow bridge and a school bus

is headed toward it, what does it do Does it crash

into the bus?

Or does it drive off the bridge, taking you

and the other passengers in the car to immediate death?

In the ancient times, we used to believe that being moral

was to transcend all your emotional responses

and come up with the perfect analysis.

But actually, an awful lot more comes

into play, our ability to read the emotions of others,

our consciousness, our understanding

of habits and rituals and the meaning of different gestures.

It's not clear that we know how to get

that kind of understanding or appreciation

into those systems.

So will their analytical tools be satisfactory?

They may win the game of Jeopardy.

And the danger of that is that will

make us liable to attribute levels or kinds of intelligence

to them that they do not have and may lead to situations

where we become increasingly reliant on them

managing tools that they won't really

know how to manage when that idiosyncratic and truly

dangerous situation arises.

I think the revolution in military robotics has been

I think just the widespread adoption and use

of the unmanned aerial systems by the US military.

As these things make mistakes, we

don't really know who's necessarily controlling them.

If they've been programmed, again, who's responsible?

So it becomes much more easy to distance yourself

from the responsibility.

And I think in the case of autonomous systems,

it's a really big question.

Because if these things accidentally

kill people or commit something which,

if it was done by a human, we would consider it a war crime,

now it's being done by a machine,

so is that just a technological error?

Or is it a war crime?

In legal terms, that's really about intention

to commit a war crime.

Otherwise, it's just sort of a mistake,

which is very different than in product liability.

So if you make a mistake with product liability,

there's a lawsuit and the company still

has to pay even though they didn't intend the harm.

But in war, that's not the case.

I think robots are not a threat in and of themselves.

I think we have to worry about how they're used.

And we have to design them and build them very responsibly.

But I think we can't be naive about the possible harms

and repercussions and the ways that they

may be used that we don't approve of or the ways that we

don't even expect, so relieving us of responsibility

for our actions because we can point to the robot

and say, well, it's not my responsibility.

I think that's a kind of abdication of responsibility.

It's a threat to who we are as humans

and how we develop as a society.

What interests me is how humans are starting

to interact with robots that views robots

not so much as objects but as lifelike things

and the ethical questions that come along with that.

Will you be my friend?

Sure thing, Martin.

Anthropomorphism is our tendency

to project human-like qualities on animals

or life-like qualities on objects.

And the interesting thing about robots, particularly social

robots, is that just the fact that we're having something

moving around in our physical space

that we can't quite anticipate lends itself

to this projection.

And we start to name these object

or give them a gender or ascribe intent or states of mind

to them.

We'll even feel bad for them if they get stuck under the couch.

And so we start to perceive these objects differently

than we do other objects like toasters.

This can get actually pretty extreme.

So I did a workshop where my friend [INAUDIBLE]

and I gave people little robotic dinosaurs

and had them play with them.

And then we asked them to torture and kill them.

And they had a lot of trouble doing it.

They basically refused to even hit the things.

So anthropomorphism can go a long way

in how we're willing to treat things.

Even with the primitive social robots we have now,

there are YouTube videos of robots being

tortured or treated violently.

And the comments underneath these videos

are very polarized.

And some people already get very upset by this type of behavior.

We're not at this stage yet in robotic development

that robots could actually experience something

that's anywhere near to how we imagine our own pain

experience.

So the question is not do the robots actually feel it

when you hurt them?

The question is more, what do we feel when we hurt them?

One could imagine especially a child who doesn't really

understand the difference between a cat

and a robotic object becoming either

traumatized or desensitized to this type of behavior.

If our research shows that violent behavior and hurting

robots translates to treating other humans in a violent way

or turning off parts of us that are empathetic,

then yes, we probably should prevent people

from hurting robots.

If we put a red line in the sand right now,

that doesn't mean that we will never

see terminator-like creatures.

But it means we're trying to design

in the appropriate ethical considerations

in today's systems.

It's going to be more important what we project onto robots

and how we interact with them and what

roles they play in society.

And that's going to drive what moral status we give them,

essentially.

And I think that's why it's important for us

collectively to also think about who we are as humans

and how we develop as a society.

[MUSIC PLAYING]