Whats it like to be a robot Leila Takayama
You only get one chance
to make a first impression,
and that’s true if you’re a robot
as well as if you’re a person.
The first time that I met
one of these robots
was at a place
called Willow Garage in 2008.
When I went to visit there,
my host walked me into the building
and we met this little guy.
He was rolling into the hallway,
came up to me, sat there,
stared blankly past me,
did nothing for a while,
rapidly spun his head around 180 degrees
and then ran away.
And that was not a great first impression.
The thing that I learned
about robots that day
is that they kind of do their own thing,
and they’re not totally aware of us.
And I think as we’re experimenting
with these possible robot futures,
we actually end up learning
a lot more about ourselves
as opposed to just these machines.
And what I learned that day
was that I had pretty high
expectations for this little dude.
He was not only supposed to be able
to navigate the physical world,
but also be able
to navigate my social world –
he’s in my space; it’s a personal robot.
wWhy didn’t it understand me?
My host explained to me,
“Well, the robot is trying
to get from point A to point B,
and you were an obstacle in his way,
so he had to replan his path,
figure out where to go,
and then get there some other way,”
which was actually
not a very efficient thing to do.
If that robot had figured out
that I was a person, not a chair,
and that I was willing
to get out of its way
if it was trying to get somewhere,
then it actually
would have been more efficient
at getting its job done
if it had bothered
to notice that I was a human
and that I have different affordances
than things like chairs and walls do.
You know, we tend to think of these robots
as being from outer space
and from the future
and from science fiction,
and while that could be true,
I’d actually like to argue
that robots are here today,
and they live and work
amongst us right now.
These are two robots that live in my home.
They vacuum the floors
and they cut the grass
every single day,
which is more than I would do
if I actually had time to do these tasks,
and they probably
do it better than I would, too.
This one actually takes care of my kitty.
Every single time
he uses the box, it cleans it,
which is not something I’m willing to do,
and it actually makes
his life better as well as mine.
And while we call these robot products –
it’s a “robot vacuum cleaner,
it’s a robot lawnmower,
it’s a robot littler box,”
I think there’s actually a bunch
of other robots hiding in plain sight
that have just become so darn useful
and so darn mundane
that we call them things
like, “dishwasher,” right?
They get new names.
They don’t get called robot anymore
because they actually
serve a purpose in our lives.
Similarly, a thermostat, right?
I know my robotics friends out there
are probably cringing
at me calling this a robot,
but it has a goal.
Its goal is to make my house
66 degrees Fahrenheit,
and it senses the world.
It knows it’s a little bit cold,
it makes a plan and then
it acts on the physical world.
It’s robotics.
Even if it might not
look like Rosie the Robot,
it’s doing something
that’s really useful in my life
so I don’t have to take care
of turning the temperature
up and down myself.
And I think these systems
live and work amongst us now,
and not only are these systems
living amongst us
but you are probably
a robot operator, too.
When you drive your car,
it feels like you are operating machinery.
You are also going
from point A to point B,
but your car probably has power steering,
it probably has automatic braking systems,
it might have an automatic transmission
and maybe even adaptive cruise control.
And while it might not be
a fully autonomous car,
it has bits of autonomy,
and they’re so useful
and they make us drive safer,
and we just sort of feel
like they’re invisible-in-use, right?
So when you’re driving your car,
you should just feel like
you’re going from one place to another.
It doesn’t feel like it’s this big thing
that you have to deal with and operate
and use these controls
because we spent so long
learning how to drive
that they’ve become
extensions of ourselves.
When you park that car
in that tight little garage space,
you know where your corners are.
And when you drive a rental car
that maybe you haven’t driven before,
it takes some time
to get used to your new robot body.
And this is also true for people
who operate other types of robots,
so I’d like to share with you
a few stories about that.
Dealing with the problem
of remote collaboration.
So, at Willow Garage
I had a coworker named Dallas,
and Dallas looked like this.
He worked from his home in Indiana
in our company in California.
He was a voice in a box
on the table in most of our meetings,
which was kind of OK
except that, you know,
if we had a really heated debate
and we didn’t like what he was saying,
we might just hang up on him.
(Laughter)
Then we might have a meeting
after that meeting
and actually make the decisions
in the hallway afterwards
when he wasn’t there anymore.
So that wasn’t so great for him.
And as a robotics company at Willow,
we had some extra
robot body parts laying around,
so Dallas and his buddy Curt
put together this thing,
which looks kind of
like Skype on a stick on wheels,
which seems like a techy, silly toy,
but really it’s probably
one of the most powerful tools
that I’ve seen ever made
for remote collaboration.
So now, if I didn’t answer
Dallas' email question,
he could literally roll into my office,
block my doorway
and ask me the question again –
(Laughter)
until I answered it.
And I’m not going to turn him off, right?
That’s kind of rude.
Not only was it good
for these one-on-one communications,
but also for just showing up
at the company all-hands meeting.
Getting your butt in that chair
and showing people that you’re present
and committed to your project
is a big deal
and can help remote collaboration a ton.
We saw this over the period
of months and then years,
not only at our company
but at others, too.
The best thing that can happen
with these systems
is that it starts to feel
like you’re just there.
It’s just you, it’s just your body,
and so people actually start
to give these things personal space.
So when you’re having a stand-up meeting,
people will stand around the space
just as they would
if you were there in person.
That’s great until
there’s breakdowns and it’s not.
People, when they first see these robots,
are like, “Wow, where’s the components?
There must be a camera over there,”
and they start poking your face.
“You’re talking too softly,
I’m going to turn up your volume,”
which is like having a coworker
walk up to you and say,
“You’re speaking too softly,
I’m going to turn up your face.”
That’s awkward and not OK,
and so we end up having to build
these new social norms
around using these systems.
Similarly, as you start
feeling like it’s your body,
you start noticing things like,
“Oh, my robot is kind of short.”
Dallas would say things to me –
he was six-foot tall –
and we would take him via robot
to cocktail parties and things like that,
as you do,
and the robot was about five-foot-tall,
which is close to my height.
And he would tell me,
“You know, people are not
really looking at me.
I feel like I’m just looking
at this sea of shoulders,
and it’s just – we need a taller robot.”
And I told him,
“Um, no.
You get to walk in my shoes for today.
You get to see what it’s like
to be on the shorter end of the spectrum.”
And he actually ended up building
a lot of empathy for that experience,
which was kind of great.
So when he’d come visit in person,
he no longer stood over me
as he was talking to me,
he would sit down
and talk to me eye to eye,
which was kind of a beautiful thing.
So we actually decided
to look at this in the laboratory
and see what others kinds of differences
things like robot height would make.
And so half of the people in our study
used a shorter robot,
half of the people in our study
used a taller robot
and we actually found
that the exact same person
who has the exact same body
and says the exact same things as someone,
is more persuasive
and perceived as being more credible
if they’re in a taller robot form.
It makes no rational sense,
but that’s why we study psychology.
And really, you know,
the way that Cliff Nass would put this
is that we’re having to deal
with these new technologies
despite the fact
that we have very old brains.
Human psychology is not changing
at the same speed that tech is
and so we’re always playing catch-up,
trying to make sense of this world
where these autonomous things
are running around.
Usually, things that talk are people,
not machines, right?
And so we breathe a lot of meaning
into things like just height of a machine,
not a person,
and attribute that
to the person using the system.
You know, this, I think,
is really important
when you’re thinking about robotics.
It’s not so much about reinventing humans,
it’s more about figuring out
how we extend ourselves, right?
And we end up using things
in ways that are sort of surprising.
So these guys can’t play pool
because the robots don’t have arms,
but they can heckle the guys
who are playing pool
and that can be an important thing
for team bonding,
which is kind of neat.
People who get really good
at operating these systems
will even do things
like make up new games,
like robot soccer
in the middle of the night,
pushing the trash cans around.
But not everyone’s good.
A lot of people have trouble
operating these systems.
This is actually a guy
who logged into the robot
and his eyeball was turned
90 degrees to the left.
He didn’t know that,
so he ended up just bashing
around the office,
running into people’s desks,
getting super embarrassed,
laughing about it –
his volume was way too high.
And this guy here
in the image is telling me,
“We need a robot mute button.”
And by that what he really meant
was we don’t want it to be so disruptive.
So as a robotics company,
we added some obstacle
avoidance to the system.
It got a little laser range finder
that could see the obstacles,
and if I as a robot operator
try to say, run into a chair,
it wouldn’t let me,
it would just plan a path around,
which seems like a good idea.
People did hit fewer obstacles
using that system, obviously,
but actually, for some of the people,
it took them a lot longer
to get through our obstacle course,
and we wanted to know why.
It turns out that there’s
this important human dimension –
a personality dimension
called locus of control,
and people who have
a strong internal locus of control,
they need to be the masters
of their own destiny –
really don’t like giving up control
to an autonomous system –
so much so that they will
fight the autonomy;
“If I want to hit that chair,
I’m going to hit that chair.”
And so they would actually suffer
from having that autonomous assistance,
which is an important thing for us to know
as we’re building increasingly
autonomous, say, cars, right?
How are different people going
to grapple with that loss of control?
It’s going to be different
depending on human dimensions.
We can’t treat humans
as if we’re just one monolithic thing.
We vary by personality, by culture,
we even vary by emotional state
moment to moment,
and being able to design these systems,
these human-robot interaction systems,
we need to take into account
the human dimensions,
not just the technological ones.
Along with a sense of control
also comes a sense of responsibility.
And if you were a robot operator
using one of these systems,
this is what the interface
would look like.
It looks a little bit like a video game,
which can be good because
that’s very familiar to people,
but it can also be bad
because it makes people feel
like it’s a video game.
We had a bunch of kids
over at Stanford play with the system
and drive the robot
around our office in Menlo Park,
and the kids started saying things like,
“10 points if you hit that guy over there.
20 points for that one.”
And they would
chase them down the hallway.
(Laughter)
I told them, “Um, those are real people.
They’re actually going to bleed
and feel pain if you hit them.”
And they’d be like, “OK, got it.”
But five minutes later,
they would be like,
“20 points for that guy over there,
he just looks like he needs to get hit.”
It’s a little bit
like “Ender’s Game,” right?
There is a real world on that other side
and I think it’s our responsibility
as people designing these interfaces
to help people remember
that there’s real consequences
to their actions
and to feel a sense of responsibility
when they’re operating
these increasingly autonomous things.
These are kind of a great example
of experimenting with one
possible robotic future,
and I think it’s pretty cool
that we can extend ourselves
and learn about the ways
that we extend ourselves
into these machines
while at the same time
being able to express our humanity
and our personality.
We also build empathy for others
in terms of being
shorter, taller, faster, slower,
and maybe even armless,
which is kind of neat.
We also build empathy
for the robots themselves.
This is one of my favorite robots.
It’s called the Tweenbot.
And this guy has a little flag that says,
“I’m trying to get
to this intersection in Manhattan,”
and it’s cute and rolls
forward, that’s it.
It doesn’t know how to build a map,
it doesn’t know how to see the world,
it just asks for help.
The nice thing about people
is that it can actually depend
upon the kindness of strangers.
It did make it across the park
to the other side of Manhattan –
which is pretty great –
just because people would pick it up
and point it in the right direction.
(Laughter)
And that’s great, right?
We’re trying to build
this human-robot world
in which we can coexist
and collaborate with one another,
and we don’t need to be fully autonomous
and just do things on our own.
We actually do things together.
And to make that happen,
we actually need help from people
like the artists and the designers,
the policy makers, the legal scholars,
psychologists, sociologists,
anthropologists –
we need more perspectives in the room
if we’re going to do the thing
that Stu Card says we should do,
which is invent the future
that we actually want to live in.
And I think we can continue to experiment
with these different
robotic futures together,
and in doing so, we will end up
learning a lot more about ourselves.
Thank you.
(Applause)