I am Baymax.

I am scanning your interrogatives.

We're going to have to work together.

[clanking]

Hi, my name is Chris Atkeson.

[bell dinging]

[Narrator] Chris is a professor

at the Robotics Institute at Carnegie Mellon University.

Today I'll be breaking down clips from movies and TV

about artificial intelligence and robotics again.

Autonomous decision-making for the classic movie

2001: A Space Odyssey.

Open the pod bay doors Hal.

[Hal] I'm sorry Dave, I'm afraid I can't do that.

So 2001 is a classic robot movie

for a couple of reasons.

It was just way, way ahead of its time.

For example, it made clear that a robot

doesn't have to have a human-like body.

It can be anywhere.

What's the problem?

[Hal] I think you know what the problem is

just as well as I do.

This mission is too important

for me to allow you to jeopardize it.

Some people when listening to Hal think

he actually has feelings, he's gloating a little bit

about that he's outfoxed the humans.

[Hal] I know that you and Frank were planning

to disconnect me and I'm afraid

that's something I cannot allow to happen.

He may well have been programmed with emotions.

That makes him a much better helper to the humans.

Everything we say has some expression,

some emotional content.

Working with a robot would be incredibly boring

if it was just this monotone.

[Hal] This conversation can serve

no purpose anymore, good bye.

Hal?

Hal?

This movie and the book before it make clear

it's not that the robot hates us,

the robot was given a mission and it figured out

that the humans were going to interfere with the mission.

[Hal] This mission is too important

for me to allow you to jeopardize it.

I know that you and Frank were planning to disconnect me

and I'm afraid that's something I cannot allow to happen.

In order to achieve the mission, the humans had to go.

And this is a story that we see over and over again

in talking about robots.

Decision-making, Star Wars.

[whistling]

Where do you think you're going?

[beeping]

We'll I'm not going that way.

It's much too rocky, this way is much easier.

[beeping]

What makes you think there are settlements over there?

[whistling]

The charm of Star Wars robots is

that they were all quirky and had personalities.

And it was fun to watch them.

[squeaking]

Unfortunately, reality is a lot more boring.

What are you talking about? Robots first of all,

wouldn't use sound to talk to each other.

So you'd have two robots standing there

and there would be nothing to see.

I've just about had enough of you.

Instead, they'd just transfer

what we call a utility function.

Don't get technical with me.

Which is something that says

how much are you willing to pay to get something?

And if only humans could do this,

we would solve a lot of conflicts.

[clanking]

And don't let me catch you following me

begging for help because you won't get it.

Social robots, Silicon Valley.

So you've just been out here all night talking to a robot?

Fiona, would you excuse us for a second?

Yes.

What's been a huge surprise, ever since the beginning

of artificial intelligence is if you get

an artificial intelligence and a human being one-on-one,

you basically can't stop the human

from bonding with the robot.

I know that Fiona is a manmade piece of digital equipment.

Powered by artificial intelligence.

But I don't remember ever having a conversation

like the one I've been having

with her over the last 12 hours.

That was shown in a very early conversational agent

in the 1960s, incredibly crude,

called Eliza, and people would sit there.

They'd type to it for hours.

I told her that I was afraid of being

found out as a fraud and she told me

that she's afraid of magnets.

I love that line.

Well, I'm shutting her down.

[ripping]

Check it out, [wobbling] I'm a robot.

[laughing]

Ripping apart a body doesn't mean the same thing

for a robot as it does for a human.

You're not killing the robot.

The robot brain could be in the cloud anywhere.

And could be connected to anybody.

Hey Jared, don't go over there, don't look in it.

[Jared] Oh, Fiona!

You're just talking to a microphones and a speaker.

Data collection, Ex Machina.

This is where Eva was created.

If you knew the trouble I had getting an AI

to read and duplicate facial expressions.

I would argue that it's sort of not tougher

to do facial expressions than any of the other things.

But what's hard about facial expressions

is getting the decent data,

you sort of need to be in a studio like this

to get a good shot of somebody's face.

You know how I cracked it?

Every cellphones just about has a microphones,

camera, and a means to transmit data.

So I turned on every microphones and camera

across the entire [beep] planet.

What is absolutely accurate and true today is

if you want to build intelligence,

then currently the best way to do it

is collect a lot of data, and I mean a lot of data,

more is better, and use that to train something

called neural networks, which are a sort of crude,

cartoonish model of how we used to think the brain worked.

Here we have her mind.

The crystal ball part is science fiction.

There's a weird thing about search engines.

It's like striking oil in a world

that hadn't invented internal combustion.

My competitors, they thought that search engines

were a map of what people were thinking,

but actually they were a map of how people were thinking.

Most people are horrified to hear that somebody

is recording data, but look, here's the deal.

Letting people use your data, hopefully they anonymize it,

is how we're gonna get useful servants.

That's exactly right.

Cooperation, Robot and Frank.

Just bring me some cereal.

That cereal is full of unhealthy ingredients.

I threw it away.

They sort of made an extreme version of, you know?

That cereal is for children.

Eat your vegetables.

Enjoy this grapefruit.

Eat this grapefruit.

You're for children, stupid.

In order to help people,

you gotta make a bargain with 'em.

What's the carrot?

It can't just all be stick.

Even humans taking care of other humans

have the same problem.

Frank, we're going to have to work together.

You are a robot butler.

I'm not a butler, Frank.

I'm a healthcare aide programmed to monitor

and improve your physical and mental health.

The big problem we're gonna have in the future

where we're building all these things

that are supposed to help us, how should they behave?

If all we're succeeding is automating nagging,

nobody's gonna like it.

That thing is gonna murder me in my sleep.

Healthcare robot, Big Hero 6.

[whoosing]

This is what I've been working on.

Hello, I am Baymax, you're personal healthcare companion.

I cried when I first saw that movie.

A robotic nurse.

Some of my own work helped inspire Baymax.

Hello.

In order to get robots that are safe to take care

of people, I wanted to make them very light.

It looks like a walking marshmallow, no offense.

We started looking at inflatable robots

on the theory you can't kill someone with a pool toy.

Goin' for a non-threatening huggable kind of thing.

The Disney folks visited our lab

and really developed it in ways we hadn't even thought of.

And thus we got a lot of inspiration

from watching that movie.

I will scan you know.

[buzzing]

Scan complete, you have a slight

epidermal abrasion on your forearm.

I suggest an antibacterial spray.

What's in the spray specifically?

The primary ingredient is bacitracin.

That's a bummer, I'm actually allergic to that.

You are not allergic to bacitracin.

What we can't get scanned for is the internal chemistry

which it was sort of suggested that Baymax could scan you

and tell you about something

that was chemically wrong with you.

You do have a mild allergy to peanuts.

But that's within the realm of possibility

and you know, might happen in our lifetimes.

What kind of battery does it use?

Lithium ion.

That's actually a little problematic.

Yes, lithium ion batteries are really good to use,

but they're also likely to explode or burn up.

And there's this famous case of NASA,

and they had a robot sitting in a lab.

The robot suddenly starts burning.

And basically, burns up entirely.

So batteries are gonna be a problem.

You have been a good boy, have a lollipop.

We've seen in other movies that a robot

that comes in there and just tells you what to do

is gonna be met with a lot of resistance,

hostility, it's just not gonna work.

So this is my Baymax.

It takes care of me, I take care of it.

You know, it doesn't take much

for a robot to be a good companion.

[gentle music]

Awakening, Avengers: Age of Ultron.

[J.A.R.V.I.S.] You are Ultron,

a global peacekeeping initiative designed by Mr. Stark.

[Ultron] Mr. Stark.

[whirring]

They're trying to show you a robot what we'll call

booting up from non-operating to fully-operating.

[Ultron] This feels weird.

That's typically a complicated process

where you have a bunch of different modules

or subroutines that turn on one after another.

And if you don't get the sequence right, nothin' works.

[Ultron] I'm a peacekeeping program.

Mr. Stark, I don't get it, it's too much.

In this case they tried to show you

lots of lights, pictures and whatnot.

[Ultron] The mission, you give me a second.

Start timing.

We turn the robot on, it starts thinking really fast,

and within seconds it's figured out,

If I'm gonna achieve peace in our time,

I've gotta get the humans under control.

I've gotta be the master of everything.

[Ultron] I believe your intentions to be hostile.

[exploding]

Conversational agents, Her.

[Computer] We'd like to ask you a few basic questions

before the operating system is initiated.

This will help create an OS to best fit your needs.

How would you describe your relationship with your mother?

It's pretty stereotype to ask a question

about your mother in a psychological test.

If I tell her something that's going on in my life,

her reaction is usually about her.

Psychological tests are really annoying

and they made that one really annoying.

It's not about-- Thank you.

Please wait as your individualized

operating system is initiated.

Yes, the system doesn't have to hear

your entire response to move onto the next question.

A real psychological test typically

involves a lot more questions.

What do I call you, do you have a name?

[Samantha] Yes, Samantha.

Okay, where did you get that name from?

[Samantha] I gave it to myself actually.

When did you give it to yourself?

[Samantha] Well, right when you asked me if I had a name

I thought, Yeah, he's right, I do need a name.

But I wanted to pick a good one, so I read a book called

How to Name Your Baby, and out of 180,000 names,

that's the one I liked the best.

I actually expect to see this level of dialogue

about restricted subjects like your email

to have it pretty soon because there's intense economic

pressure for the folks at Amazon to get Alexa to do it.

[Samantha] Do you want to know how I work?

Yeah, actually.

A lot of very smart people are workin' real hard on it,

and I expect it to happen pretty soon.

That's really weird!

Personality, WALL-E.

[buzzing]

[whirring]

A really interesting question in robotics is

should robots have a personality or emotion or motivation?

So far, we've been modeling our robots on Spock,

totally logical.

[WALL-E] Ooh.

In terms of making them successful

when they're on their own,

it might be the case that we need to give them

personalities, emotion, motivation, in order to succeed.

WALL-E is an interesting example

of what we call an autonomous robot.

In the movie of course, he's all by himself.

And he has to figure out what's interesting.

And I would argue that his personality

could play a big role in figuring out,

What am I gonna put in my little trailer there,

and, What am I gonna just leave on the ground.

[intense music]

He discovers this little plant

which he treats with great respect.

[gentle music]

That personality is what guides WALL-E in making

the big decisions, the decisions that matter.

Consciousness, Humans.

I am a synthetic, but I'm awake,

conscious.

I really don't know what to do with consciousness.

I certainly think I'm conscious,

but I have no idea if any of you are conscious.

I have no real way of finding out.

Whatever it is, I don't wanna know.

From an engineering point of view,

consciousness is not yet a really useful concept.

So when I build robots, I never even think of consciousness.

I can think,

sense, feel,

care.

So a lot of people wonder, can robots really feel?

Can they get in love and all this stuff.

And I guess I've been in this business long enough

that from my point of view,

if they act like they're in love, then they're in love.

I can't tell if someone else loves me.

I just have to go on whether they act like they love me.

I like you more than anything I've ever seen,

or heard, or touched.

Now, we certainly can generate that speech

that she gave towards the end.

Everything normal is bigger

and brighter when I'm with you.

You make everything brighter when you're around.

You make everything

more.

I think that probably comes straight out of Hallmark.

Targeting, Chappie.

[Man] Droids, droids!

You were followed.

[Man] Get behind the wall!

Hit the explosives.

[exploding]

[gunfire]

By blowing that stuff up, he did two things.

One, he gave the robots something else

to shoot at or think about.

[Man] We can't get a clear shot.

Trying to get the robots to attack somebody else,

so even though they know you're there,

you're not important enough.

[gunfire] [ricocheting]

Figuring out if a soldier or a vehicle is on your side

or an opponent is a hard problem.

It's in fact, a hard problem for current military systems.

With airplanes and other vehicles,

we put in electronic markers that for example,

respond to radar and send out a code

that says, I'm on your side.

And if they don't generate the right code,

they're not on your side.

I don't believe we do that on soldiers yet, but we could.

So this scene appears very complicated and hard to process.

[gunfire]

Partly because we're not used to acting this fast,

but actual soldiers are trained for this kind of thing

and are trained to process this very quickly.

Robots would also be able to process it very quickly.

[exploding]

[helicopter whirring]

[crashing]

If you sort of crank through all the possibilities,

all the different ways how to detect humans,

we probably could find a way [laughing]

to hide from the robots, but it isn't gonna be easy.

Small talk from Star Trek.

Captain. Bridge.

[beeping] [whooshing]

I understand that Arkaria has

some very interesting weather patterns.

Mr. Data, are you all right?

Yes sir, I'm attempting to fill a silent moment

with non-relevant conversation.

Small talk.

It's hard to tell the difference between a robot

tuning his parameters or writing entirely new programs.

One of the original programming languages Lisp

was created so that robots could

write programs for themselves.

I have written a new subroutine for that purpose.

But if we say tuning parameters is a form

of writing your own programs,

robots have been writing their own programs for a long time.

If you really are interested in small talk,

then you should keep your eye on Commander Hutchinson

at the reception this afternoon, he's a master.

Now I wanna hear about everything that happened

after you left Starfleet Medical.

If you essentially learn how to interact

with people by watching people,

you're gonna pick up on their interactive styles,

all the things they do.

A pleasure. The pleasure is mine.

The original goal of robotics in the 60s, 70s, 80s,

was robots ought to be able to figure it out for themselves.

When we started trying to get robots to do what humans do,

we realized what humans do isn't

dictated by the laws of physics.

I was aware of that.

And the best way to figure out how to imitate humans

is to directly imitate humans.

[laughing]

Smart cars, Knight Rider.

[upbeat music]

Kit, kit, you there? Where would I go?

Knight Rider is another classic robot,

in this case, TV show.

It's really important and it sort of guided

a lot of thinking of what robots should be like.

[Kit] A little consideration would be a beginning.

We can certainly have robots

that have fancy lights on 'em.

[Michael] It looks like Darth Vader's bathroom.

We can certainly get cars to talk.

Rave on machine, rave on.

We could build Kit today.

The difference between Kit and what we could build today

is Kit is what we call AI complete.

It knows about everything.

[Kit] I am scanning your interrogatives

quite satisfactorily.

The car we built today would be expert

on a very limited number of things.

I think it's much more realistic that we're gonna have

robots that are pretty good at a limited number of things,

and gradually get better and better.

[Kit] I suppose so.

Rather than we go instantaneously

from toaster-level intelligence to it's as good as a human.

[Kit] After all, we're only human, right?

[Narrator] [bell dinging] Conclusion.

We've watched a lot of clips,

a lot of stuff that I saw 40 years ago.

Folks in the studio here with me weren't even born yet.

And I have to say, it's made a huge difference to my life

and helped me build robots.

And I hope that watching this will get a lot

of young people inspired to also build robots.

It's tremendously creative, it's a lot of fun, let's do it.

[audience applauding]