"How to Survive a Robot Uprising" - Daniel H. Wilson...


Uploaded by Google on 16.07.2007

Transcript:

MALE PRESENTER: Welcome, everyone, today, for our
authors at Google presentation by Daniel Wilson, who is a
recent PhD graduate from the Robotics Institute at Carnegie
Mellon University, where he also earned his master's
degree in data mining.
Daniel has worked at various top research laboratories
including Microsoft Research, Xerox Palo Alto Research
Center, and Intel Research.
Currently Daniel is columnist for Popular Mechanics, and How
to Survive a Robot Uprising is his first book.
The movie rights were recently optioned by Paramount Pictures
with Reno 911 writers set to write the script.
He's currently working on his second book, tentatively
titled Where's My Jet Pack?
We're thrilled to welcome Daniel to Google today, and
would like to thank him for making a special
trip to visit us.
[APPLAUSE]
DANIEL H. WILSON: Hi, everybody.
Yeah, thanks for coming out.
This is great.
What's especially great is Google already bought all
those books.
So I don't actually have to convince you to buy them.
Yeah, this is what you're supposed to do once you get a
PhD in robotics, I think, is go on a book tour.
So, yeah.
My name's Daniel, and I wrote this book called How to
Survive a Robot Uprising.
And really, the title of the book says it all.
So this is a book that gives advice to humans on how to
survive when the robots inevitably come and try to
kill us all.
Now if you haven't sort of caught on
yet, the book's funny.
Yeah.
And it says humor on the back, too.
So that's usually the first thing I say during a talk.
The book's funny.
Unfortunately, since the book was published, I actually get
lots of emails from people that sort of disagree.
And it's really interesting to think about the real robot
uprising, but I sort of have to say up front that, and this
may be disappointing to you all, but I don't actually
truly believe that a robot uprising is on
the horizon for us.
And I sort of say that from an informed standpoint.

I have a little card that says my PhD is on it, I usually
wave that at people.
I'm working on getting a badge for it.
But I'm going to say to the people that are worried about
it, I don't know what to tell you except, just
buy the damn book.
So that's my advice.
Also you'll notice, recently I came up with an idea.
The book has got a bit of a foil cover, and this isn't
scientifically proven, but I'm pretty sure that if you were
to line the walls of your basement with it--
[LAUGHTER]
DANIEL H. WILSON: --the robots, they can't find you.
But that's not scientifically proven.
But it's your safety.
So what's it worth to you?
[LAUGHTER]
DANIEL H. WILSON: That's all I'm going to say.
So now I want to follow this up by basically saying that,
although I don't believe a robot uprising is really going
to happen, all the information in the book, every robot that
I talk about, every sensor that I talk about, all the
problems that I talk about that robots have to solve in
order to function, all of that is real.
And so I did spend the last five years getting a PhD in
this stuff, and although that doesn't make me an expert in
everything, I do, luckily, have lots of friends who are
really smart.
And so everything and all the advice in the book actually
comes from real roboticists.
So that's all the people at CMU, mostly.
So I'll talk a little bit about my background.
Right, I wrote this book in my spare time while I
was getting a PhD.
I also got a master's in robotics and a master's in
data mining, whatever that is.
And while I was at CMU, over time I started to notice that,
especially in America, when you're watching TV or you're
watching a movie and there's a robot, a lot of times they
start out good, but it always ends with human limbs flying
in the air, and people getting killed, and the red eyes, and
the spinning heads.
Over time, I became more and more irked by this until it
finally sort of culminated one night at a bar.
And I was talking with my friends and I was kind of
griping about this, and then I realized suddenly that, like
all people must at some point in their lives, I was at a
point where I was qualified to write a book called How to
Survive a Robot Uprising.
I realized that I could actually seriously respond to
all these uprising scenarios that everyone
is so familiar with.
And for most people, I think that's all they know about
robots really, are what they see in movies.
And so I wrote the book sort of to strike back at
Hollywood, to sort of take them seriously, get it all
down, and just answer all the questions for good.
And call out all the stereotypes.
So I really liked the idea for that reason.
So after I thought of it, the first thing I did was I wrote
a lot of jokes.
Like jokes about how to ingratiate yourself to robot
overlords in a post-apocalyptic world.
How to make love to robots, things like that.
Crucial information granted, but not
necessarily based on fact.
So the first thing that my editor did when he saw the
sort of initial manuscript that I handed him, was he took
a blue pencil and he sort of crossed all of that stuff out.
And in retrospect, I kind of agree with that.
The idea is that with robotics, a lot of times the
fact is stranger than the fiction.
So everything in the book is really rooted in fact.
And so what I realized I needed to do was, I had to go
around and ask roboticists about how they would deal with
each of these situations.
So if we're talking about a terminator, well, there are
people that work on humanoid robots.
All aspects.
There's bipedal locomotion and vision and things like that.
If we're talking about escaping from an unmanned
ground vehicle, there are lots of people that are working on
making cars that drive without people.
So, personally, my research is just in smart homes that track
the elderly.
So although that's going to be really useful, I mean, those
robots, most dangerous, hands down.
So I know a lot about that.
So I had this one section that I could write, How to Escape
From a Smart Home.
And I really had a lot of fun with that section, and it's
actually one of the longest sections because I just had so
many jokes had to get in there.
But what about all the rest of the book?
So I went around CMU, and I had a lot of fun.
My adviser is actually a guy who is a humanoid robotics
guy, and so I just sort of, after an adviser-advisee
meeting one day, I kind of paused and said, so, Chris.
Let's just say Terminator is chasing you.
[LAUGHTER]
DANIEL H. WILSON: What would you do?
And he kind of looks at me.
What are you talking about?
And then right about then, a post-doc I was sharing an
office with was like, well, what kind of
humanoid robot is it?
What kind of terminator?
And then it goes from there.
And that's it.
And I was really amazed by how game everybody
was to help me out.
Once you get them started--

how big can a bipedal walker get?
Is it like an imperial walker?
Is that feasible?
Well, it would have to be passive dynamic, and just goes
and I'm just like scribbling down notes.
Yes, aha-ha.

And actually I also found something else that was kind
of interesting from talking to all these roboticists who
varied in age and expertise.
I don't name names or anything in the book.
I have an Acknowledgements section.
But I talked a lot with graduate students.
I was just interested in getting all the
creepy little stories.

But what I found was that, in one section, I had trouble.
Oh, sorry I forgot to mention.
The one thing I found it was interesting was that whenever
you ask these people why they're doing their jobs,
generally the reason is because they
think robots are cool.

I really saw that they thought robots were cool whenever they
start talking about them eviscerating people,
ironically.
That's whenever it really shined through.
That was kind of nice to see because I sort of thought
maybe I was the only one who saw an episode of Small Wonder
and then decided to get a PhD.
Anyway, so yeah, I did have trouble with one section.
And this is a section that you want to take
with a grain of salt.
I did my best. It's called How to Treat a Laser Wound.
And this, I feel because this could be crucial at some point
in the near future.
But I had to actually call physicians for this.
As it turned out, every place I called, every burn unit I
called, no one would give me advice on the record because
they were worried about being sued over this, which sort of
leads me to, if I just go with that, then what this means is
we're going to have somebody who buys the book, gets
attacked by like a man-eating robot, gets a limb maybe loped
off by by a laser, and then comes back and sues me.
I'm into it.
That's going to be really fun when that happens.
So anyway, that's just from an army field
manual, that section.
So what I'm going to do is, I'm going to hold up my book
as I was instructed to do and forgot.
And I'm actually going to read from the book
for about ten minutes.
But before I do, I'm going to tell you what kind of advice
is in this book.
So first of all, some of the advice is just plain funny.
It's just jokes that I thought of.
For instance, you should be suspicious if your servant
robot is constantly talking about human-killing and making
repetitive stabbing movements.
So as it turns out, I didn't ask somebody about that.
I just made up and wrote it.
But still good advice.
And did I mention I've got a PhD in robotics yet?
Yeah, PhD, kids.
So anyway, some of it's also taken straight from
roboticists.
So these creepy little facts.
So there's a guy named Metin Sitti--
I hope I pronounced his name right--
that's working at CMU right now.
And he makes these robotic geckos and robotic flies.
As it turns out, a robotic fly, it looks like sort of a
pencil eraser with wings.
But it will sound exactly like a real fly, and that's because
it beats its wings at the same frequency.

It mimics the way a real fly's wings work.
And that's kind of creepy, right?
So now every time you hear a fly, you know it could be
robot fly, could be a real fly, I mean there's definitely
a probability that leans towards real fly on that one.
I'm going to go on a limb and say that.
And finally, the sort of the third flavor of advice, is
just extrapolated advice.
So, for instance, I have a section on hand to hand combat
with humanoid robots.
And so far as far as I know there's not a robot that
exists really just to sort of spar with a human, just to
sort of beat them about the head and neck and face.
So what I had to do is I kind of had to find a humanoid
roboticist and ask about how the arms work and how fast
they could throw a punch and things like that, and
extrapolate that in order to get the
advice for that section.
So that said, I'm now going to read from the book for a
little while.
I'm going to start with the introduction.

If popular culture has taught us anything, it's that someday
mankind must face and destroy the growing robot menace.
In print and on the big screen, we've been deluged
with scenarios of robot malfunction, misuse, and
outright rebellion.
Robots have descended on us from outer space, escaped from
top secret laboratories, and even traveled back in time to
destroy us.
The cultural icon of the killer robot goes back almost
as far as the notion of the mad scientists who supposedly
create them.
Even the word robot has ominous roots.
It's Czech for laborer and was coined in R.U.R., Rossum's
Universal Robots, a play produced in 1920 in which
robots revolted and destroyed all humans.
Who would have thought?
Today scientists are working hard to bring these artificial
creations to life.
In Japan, fuzzy little Real Robots are delivering much
appreciated hug therapy to the elderly.
Children are frolicking with smiling robot toys.
It all seems so innocuous.
And yet, how could so many Hollywood scripts be wrong?
I ask you.
[LAUGHTER]
DANIEL H. WILSON: Ask yourself that.
How could millions of dollars of special
effects lead us astray?
So take no chances.
Arm yourself with expert knowledge.
For the sake of humanity, listen to serious advice from
real robotics experts.
How else will you survive the inevitable future in which
robots rebel against their human masters?

OK, now I'm going to read a boring part.
I'm going to read an informative part.
And I also like to read this part because it contains my
favorite typo of the book.
And it's a doozy, man.
I love it.
This is Robot Sensors.
Robots are unlike any adversary
heretofore known to man.
They will use any means available to sense and make
sense of the outside world.
We cannot even imagine the scope and depth of the
information available to them.
Though we can roughly define their sensors in terms of
human abilities, robots are truly superhuman.
A sensor is any device that converts a property of the
physical world into an electrical signal.
The five human senses are visions, hearing--
yeah, did you get that?
Visions?
But, honestly, I'm just standing by it because it's
written down.
It's got some credence.
Visions, hearing, touch, smell, and taste.
Robots have a much wider variety of sensors to choose
from, each of which supplies different information and has
its own particular vulnerability.
What matters to us is whether a sensor is visible or hidden.
Extrinsic sensors inform a robot about the outside world,
and are vulnerable because they are usually placed on the
outside of the robot.
Intrinsic sensors monitor the robot's internal state and may
be well-protected, often place deep within the robot.
Passive sensors watch quietly without changing the
environment.
Active sensors, like the sonar ping from a submarine,
aggressively inspect the environment.
Active sensors may collect more information, but they can
also give away the position of the robot.
Robots are tough, but their sensors are usually fragile.
They can be damaged when exposed to extreme
temperatures, vibrations, moisture,
thermal shock, or corrosion.
When a sensor receives too much stimulus, it can become
saturated and cease to function.
Mishandling a sensor can cause it to detect things that don't
really exist, a false positive, or to miss things
that are really there, a false negative.
In this section we'll examine those common sensors used by
robots, starting by grouping them into the five human
senses, and then exploring past human capabilities and
into the realm of super human sensory ability.
And I'm going to skip that section now and get right to
some crucial advice, which is how to survive hand to hand
combat, which is, I think, why you're all here today.
How to Survive Hand to Hand Combat.
If you find yourself in a brawl with a robot, your only
hope is to escape.
A robot foe won't trade insults and it can't be
intimidated.
You should fully expect a swift pincher clamping attack
without warning.
Follow the rules of disengagement.
Every second you spend within arm's reach of a robot can
take years off of your life.
All of them.
Destroy or disable exposed sensors.
Sensors are by far the most vulnerable
exposed part of any robot.
Destroy or disable outward facing
censors such as cameras.
A handful of dirt, mud, or water will suffice.
It's hard for a robot to wipe mud from his eyes when it has
whirring buzz saws for hands.
[LAUGHTER]
DANIEL H. WILSON: That's a little technical tidbit that I
picked up around the FRC.
Keep your hair short and your clothes tight.
To consider the alternative, imagine getting a haircut in
the garbage disposal.
Don't bother with karate unless you can punch through
sheet metal.
And find a weapon.
Your pathetic human hands are useless here.
Choose a blunt or pointed instrument.
Serrated edges don't work against
metal or durable plastic.
Even a simple crowbar can save your life.
You can run away while the robot condescendingly bends it
into a pretzel shape.
Keep your distance.
A humanoid robot can block or throw a punch about twice as
fast as a human black belt can.
In comparison, a typical inebriated human brawler
doesn't have a fighting chance.
And finally, get away.
Pretend that you just lit the fuse on a cheap Chinese
firecracker the size of the doghouse.

That technique really works whenever you need to scamper.
So now I'm going to move on to another bit of advice called
How to Deactivate a Rebel Servant Robot.
So you've discovered that your extremely submissive, lovable,
and expensive servant robot has turned rebel.
This can feel like losing a member of the family.
However, if the situation is not dealt with properly, it
may feel more like losing every member of the family.
Plus a few neighborhood kids.
Pretend everything is normal.
To forestall a mechanized killing spree you must act as
though nothing is amiss.
When your servant hands you an old tire half-full of
rainwater and mosquito larvae instead of an iced tea, simply
sip politely, nod, and smile.
That's my favorite joke in the whole book.
What does that say about me?
Mosquito larvae.
[KISSING SOUND]
DANIEL H. WILSON: OK, sorry.
That's terrible.
Send the robot on an arduous task.
Not only will sending your robot on a long, tiring task
drain its power reserves, it will give you time to
formulate a plan.
Formulate a plan.
Call the cops.
The most straightforward solution is
also the most costly.
A confrontation with law enforcement officers will
likely end with your house and servant
resembling Swiss cheese.
That's my least favorite joke in the book, actually.
Only call the fuzz in a bona fide emergency or if you have
an extremely reasonable malfunctioning killer robot
insurance deductible.
The Power Drain Plan.
Instruct the servant robot to clean the house, landscape the
yard, and assemble several major
pieces of Ikea furniture.
Then, when your robot is power-depleted and attempts to
recharge, just shut off the power to your house.
Now simply wait until the robot runs out of batteries.
If it tries to move, apply pressure with a crowbar.
The Pool Ruse.
Use this trick if you have a swimming pool.
Throw a handful of leaves into the pool and ask your loyal
robot to fetch them by hand.
When it leans over, plant your foot on its metal hind
quarters and shove.
If your robot is a waterproof model, use the next few
minutes to run away screaming.
Purchase a new manual kill switch.
You should harbor no doubts now about shelling out for a
reinforced encrypted manual kill switch, complete with a
fist-size cherry-red button.

OK, I'm going to read one that's slightly less funny in
my opinion, and slightly too serious.
This one's called How to Fire a Weapon at a Robot.
Without the proper preparation, firing a gun at a
robot can be as effective as holding the
barrel to your own head.
Robots are capable of tracking bullets to their origin either
acoustically or visually.
How then, to realize your dream of spraying a robot with
hundreds of rounds from a standard assault rifle?
Stay alert.
Watch for microphone arrays or camera arrays that point in
360 degrees.
They won't necessarily be mounted on the robot
you're firing at.
They may be spread throughout the environment or separately
mounted on several robots.
Keep moving.
Never fire from a static position.
A robot might return fire to your exact location within
milliseconds.
Try to make sure it fires where you were, and
not where you are.
Coordinate your fire with comrades.
Spread out around the target and begin firing
simultaneously.
Incoming fire from multiple directions may negate the
robot's bullet tracking.
At the very least it will make it less accurate.
Choose a complex environment.
Waterfalls, street traffic, or adverse weather conditions can
drown out the clues that robots use to
pinpoint your position.
Enclosed environments with many obstacles and surfaces
can muffle and reflect sounds, further
concealing your firing position.
And finally, modify your weapon's acoustic signature.
When you fire a weapon, a robot's acoustic bullet
tracker listens to the sound vibrations from the muzzle
blast and the supersonic crack as the bullet speeds along.
A silencer can foil some acoustic detectors.
OK.

Let's see now.
OK, so finally I'm going to read one that's
probably too funny.
And this is another crucial section called How to Pose as
a Humanoid Robot.
During an infiltration or escape, you'll need to pass
unnoticed by robot surveillance.
Most robots will be readily identifiable to each other
through encrypted markers.
How will you convince the robots that you are warm
circuits wrapped in a thin candy shell?
Pretend to be damaged.
A damaged robot may exhibit strange behavior while failing
to transmit identification.
Change your heat signature.
Stuff aluminum foil in your pants.
Rub your exposed skin with cool mud.
Hang a hulking piece of gold medal around your neck and
slip into an into an Adidas jumpsuit.
I mean, that's just good advice.
[LAUGHTER]
DANIEL H. WILSON: Your heat signature will not match a
healthy robot, nor will it match a healthy human being.
Make some a noise.
An occasional screeching beep or boop should suffice.
Make it quick and strangled.
This is no audition.
Finally, move like a robot.
Early robots exhibited a trademark clumsiness that
spawned a dance called the Robot.
Contemporary robots are more dexterous,
unless they're broken.
Pretend you're either damaged machinery or a well-oiled
break dancing machine, and pop and lock your way into the
heart of robot territory.
You guys already know how to pop and lock, right?
Because that's a whole other book if you don't.
And possibly a video.
If confronted, keep moving and don't look back.
You're just a poser, so ignore other robots and pretend to be
completely oblivious to the environment.
Keep your head down and shuffle forward with a
steady, even pace.
The fate of the entire human race may depend on it.
And finally, I'm going to read the introduction to the last
section of the book, which should get you guys all riled
up and ready to fight some robots.
Or not.
Silicon Versus Grey Matter: Winner Takes Planet.
We may have won a few battles, but humankind
must win the war.
Most likely, the epic struggle of man versus robot will not
be fought by soldiers on a smoky battlefield.
It will be acted out by average men and women and
their unruly appliances.
This is the morning that you wake up and your toast is not
made, your house is not cleaned, and your television
only shows static.
Outside the window, robotic lawn mowers are chasing people
down the street.
Inside the house, the vacuum cleaner is eyeing you angrily.
At last we turn to the purpose of this book, and the plot of
a thousand doomsday science fiction stories: how to
survive a robot uprising.
There are many potential causes of
a mass robot uprising.
A programming mistake, mistreatment by humans, or
lust for gold.
One thing is for certain.
A robot uprising will affect every person in an
industrialized nation.
Wherever there are people who enjoy purchasing time saving
gadgets at low, low prices, there will be
robots to serve them.
Densely populated city centers will be hard hit, and the
peaceful suburbs will be overrun.
Paved roads and sidewalks that allow access to humans with
disabilities will also accommodate the
wheeled robot masses.
The twisting dirt paths of the wilderness may offer natural
resistance, but as we know robots can invade any domain,
however inhospitable.
When it arrives, the robot uprising will be a coordinated
war between the two greatest
intelligences is on the planet.
Human survival hinges on our alertness to the
growing robot threat.
The robots that we use daily, those we may even call our
pets, friends, or lovers--

or all three?
I don't know if there's maybe a market for that-- will turn
on us eventually.
There may be months of meticulous planning, or we may
face a sudden, unexpected mechanical maelstrom.
The time before the inevitable attack, measured in months or
in minutes, must be a time of vigilance.
When the robot uprising begins, there will be no time
left to memorize the lessons in this book.
And now I'm done reading to you guys,
and I'll take questions.
[APPLAUSE]
DANIEL H. WILSON: Thanks.

Yes, sir?
AUDIENCE: I've actually got two.
First off, looking through the table of contents, you don't
seem to address the aspects of possibly
fighting back legally.
I mean, these are violating the Three Laws of Robotics.
Have you considered lawyers and suing them and anything?
DANIEL H. WILSON: I haven't been considering
them as legal entities.
I've mostly been considering them as insane lawn mowers.
Waking up with a Roomba on your face.
[LAUGHTER]
DANIEL H. WILSON: Which I've decided finally my advice is
to just let it finish.
But is that gross?
Yeah, that's gross.
AUDIENCE: Let's assume that you're leading a rag-tag
Colonial fleet after a Cylon uprising.
DANIEL H. WILSON: This is all stuff that got
blue penciled, but--
[LAUGHTER]
AUDIENCE: What would your strategy be in that case?
DANIEL H. WILSON: Wait, are you saying if you are on the
planet still and it'a being run by--
AUDIENCE: Let's say you're Adama.
DANIEL H. WILSON: Let's say I'm what?
AUDIENCE: Adama.
Commander of the rag-tag Colonial fleet
in Battlestar Galactica.
DANIEL H. WILSON: Oh, oh, oh, oh.
[LAUGHTER]
DANIEL H. WILSON: Well, yeah, sorry.
I'm a little lacking in the Battlestar.
OK, so if you're on the planet, I'm going to say the
first thing you want to do is you want to salvage all the
encyclopedias and things like that for the post-apocalypse.
Because, you know, predicting eclipses is a really good way
to get chicks.
But there's that.
Sorry, I'm taking from all the stuff that got blue penciled
and I'm mentally trying to get it out.
So I would say, just give it up.
Go find a new planet.
AUDIENCE: You don't seem to be very interested in the
opportunities of good life, of being a quisling and assisting
in the uprising for the potential
of continued survival.
I was wondering why you're not advocating
some of those positions.
DANIEL H. WILSON: So my office at CMU is across the hall from
Hans Moravec, and mostly just to piss him off, I think.
No.
So I do have a section in the book on how to sort of enhance
yourself and to sort of acquire all the attributes
that the robots have. Because, I mean, we invented them.
We should have them, too.
I want another arm, right?
Everybody?
So I do talk about that.
But in terms of turning traitor, you got the wrong
book, mister.
This is all about humans.
[LAUGHTER]
AUDIENCE: So we're carrying them everywhere we go.
We feed them electricity.
We stare at their screens and take their instructions.
Haven't they already won?
[LAUGHTER]
DANIEL H. WILSON: Damn it.
[LAUGHTER]
DANIEL H. WILSON: No, actually, yeah, there are
varying levels of acceptance, right?
And so you may want to Google for a neo-luddite I think.
[LAUGHTER]
DANIEL H. WILSON: You find some friends.
AUDIENCE: I had two questions, but he stole my
robot lackey question.
So I would like to ask, what do you think is the most
accurate depiction in Hollywood of the upcoming
robot apocalypse?
DANIEL H. WILSON: Yeah, so in the book I kind of describe a
time line for a robot uprising.
And I actually do make fun of several movies.
And I give one my seal of approval.
And that's I, Robot.

I don't give it the seal of approval because it's my
favorite movie, but the chain of events really
makes a lot of sense.
Like there's some sort of central AI
that's really smart.
There's a a lot of consumer robots out there that people
have. They catch a virus.
It makes them do bad things.
This really makes a lot of sense to me.
Also my other favorite part, I know people are raising their
hands, but my other favorite part about I, Robot is that
when the robots turn bad, they glow red.
[LAUGHTER]
DANIEL H. WILSON: No, that's a key design feature.
[LAUGHTER]
DANIEL H. WILSON: If you're in the kitchen and you're like,
hey robot, make me some toast. And the toast turns around and
it's glowing red, you're like, no, it's cool.
[LAUGHTER]
DANIEL H. WILSON: Bless that roboticist who added the LED
because otherwise it's just all attack and everything.
No warning.
Yes, sir?
AUDIENCE: So when I go home I should check my Roomba to see
if it's glowing red.
No, but the question I was thinking about is, aren't you
already a traitor, a quisling to the human race, by
writing this book?
Won't robots read it and then know--
[LAUGHTER]
DANIEL H. WILSON: Well, actually that is sort of part
of the irony of the book, is that when I was interviewing
roboticists.
about what are the weak points of the robots, that was always
exactly the same as their research focus.
Because that's where they're showing up deficits every day.
So anybody that works in robots, from that point of
view, is going to be a traitor.
AUDIENCE: Or is this misinformation?
Are you trying to mess with us by writing this book and the
real strategies will be passed orally human to human?
DANIEL H. WILSON: I don't know.
I'll tell you the truth.
It was just ghostwritten.
The computer wrote the whole thing and I woke up one day--
[LAUGHTER]
DANIEL H. WILSON: --and it was like, look, I sold this book,
just put your name on it with your big, fancy PhD.
[LAUGHTER]
DANIEL H. WILSON: Sorry.
AUDIENCE: So the section on cockroach-inspired robots.
You have a story.
You mentioned that one graduate student has already--
DANIEL H. WILSON: Yeah.
AUDIENCE: Like what's wrong?
Are there any other good stories of mishaps and robots
gone bad already?
DANIEL H. WILSON: Yeah.
So I really like that story when my friend
showed me that story.
So I go in-- this is RISE, R-HISE.
There's some really complicated word that means to
climb, and this is what this robot is designed to do.
And so I go and he's like, check out this robot.
He has a big briefcase, he flips it open.
It's got the foam cut out with this like upside down hexapod
like with these legs and everything.
And I look at it, and on each foot it's got a titanium
spike, right.
Oh, sorry.
And then he turns to tell me this story, oh yeah, by the
way, and I'm like no, no, no, no, no.
Let me guess.
It's got six titanium spikes on it.
This is not a huge surprise that it tried to climb a human
and he was attacked.
So sometimes the stories aren't such a surprise.
But every time I could get a good story like that, I put it
in the book.
And actually, I became a little bit of a connoisseur
about this because normally in a PhD you're trained to kind
of ask the mean questions.
Oh, that thing's amazing.
I bet it dies after five minutes.
What's the battery life on it?
Or how long before breaks?
But then whenever you're trying to write a book and
find out interesting things, it's a whole other set of our
questions you ask.
And so I'm doing this column for Popular Mechanics, and I
recently interviewed this guy at MIT who is building,
basically, biologically inspired muscles, actuators
for robots.
It's just a piece of plastic, basically, that will convulse
whenever you run current through it.

This thing exists in a liquid solution.
So I say, well, when it actuates does it squish?
Are we going to have like squishy robots?
And this is a question you would never ask as a
roboticist, but as of writing a column.
And the guy says, what?
Why do you want to know this?
What is this?
What do you mean squish?
It makes a liquidy sound.
And I'm like, oh, it's squishy.
And he's like, no, stop.
[LAUGHTER]
DANIEL H. WILSON: And so, yeah.
It's really fun to put on that other hat then be like, what's
it sound like?
What's it smell like?
Can it cut my head off?
AUDIENCE: So in addition to all these movies about
fighting robots, there are all these great movies that were
made about fighting Communists.
Like Red Dawn.
I was wondering if any of those techniques might be
applicable to fighting robots.
Is that basically the difference between fighting
robots and fighting Communists?
[LAUGHTER]
DANIEL H. WILSON: I think that this is sort of an
interchangeable.
You can interchange the word Communist with aliens.
And also zombies.

Maybe werewolves.
Pirates.
There's just so many possibilities out there that I
can't believe they made Herbie II.
[LAUGHTER]
DANIEL H. WILSON: Actually, I'm sorry.
The screenwriters that wrote the screenplay for this
possible movie based on this book, they wrote Herbie I,
which should give you guys a lot of confidence.
It does me.
I can't wait.

AUDIENCE: Which robot has most contributed to your personal
conviction that we're not really in danger?
DANIEL H. WILSON: Oh, sorry, I should be repeating these.

Which robot has most contributed to my feelings
that robots are not going to be a huge danger to people?
Pretty much, that's every robot.
Very few robots actually do what they're supposed to do
twice in a row.
So that first time is really easy, and hopefully they'll
give you your degree.
Anyway.
So actually every time I interviewed people and I would
put them in this situation, there would be a little bit of
decompression time where they kind of get into the mindset
of, OK, what if robots really were dangerous?
What would I do?
But at the beginning, roboticists would always say,
oh, if I wanted to escape from my robot, why
I'd walk away slowly.
[LAUGHTER]
DANIEL H. WILSON: Or I'd go up one stair.
Because the damn thing won't come up one stair.

Probably the most advanced robot in terms of being all
put together and doing a lot of different things at CMU is
they've got one of the Honda ASIMOs, which is a little
humanoid robot, looks like an astronaut.
And it's actually so expensive that it's got like this big
chain, and it's like, it does this and then it's like, oh,
it's going to fall, oh no.
And they baby it so much.
They're constantly afraid.
It's in a basement somewhere.
And so being face to face with real robots is not generally
that inspiring.
But then again, every now and then.

I'm really impressed with the advances made in the DARPA
Grand Challenge.
Like that amazed me that they actually did stuff.
So every now and then you see something and you go, oh my
God, we're actually making progress, which is nice.
Finally.
That's also sort of the seed of this other book
Where's My Jet Pack?
Where the hell is it?
Where's my servant robot?
AUDIENCE: So as for robots reading this book, it's
covered in foil so I think we're OK.
DANIEL H. WILSON: Right.
AUDIENCE: They're not going to be able to find it.
But are there any tips on deciding whether or not a
robot might be trustworthy?
Maybe he doesn't want to rebel.
Maybe he's on our side, so to speak.
DANIEL H. WILSON: Yeah.
Well, so I have that section on, just before How to
Deactivate a Servant Robot I have How to Spot a Rebel
Servant Robot.
So how to determine if a robot's like gone over to the
other side.

In a serious sense, a lot of that just has to do with
whether the robot's doing what it's designed to do.
And right now, what we mostly see are robots that are
designed to do one thing.
So like your Roomba.
Let's pretend you're a Roomba and you
decide, I'm sick of this.
I'm killing everybody.
And then you're like, wait a minute.
I'm three inches tall.
I'm a circular disk covered with suction cups.
[LAUGHTER]
DANIEL H. WILSON: I mean, I just imagine that's what it
would think.

So it can be difficult.
And honestly, I totally forgot the question at this point.
[LAUGHTER]
DANIEL H. WILSON: Better not to hide it.
Oh, actually, ironically, that's what I was
going to say next.
State transparency.
That's the other thing you want to have. You want to know
exactly what's going on with the internal
state of the robot.
That makes it much safer.
Hence the red light in the chest.
AUDIENCE: Given the persistent rumors that we're developing
Skynet, did you feel frightened by our invitation
to come here?

DANIEL H. WILSON: No, OK.
Because you guys, I think we've got some common ground.
Now I gave a talk, actually I gave a terrible, terrible book
talk, in Berkeley.
And everyone was very serious about this.
And they really want to know who to trust and everything.
And that was not my area so much.
I like telling jokes about robots.
I like pretending to be a robot and then narrating that
experience like I just did.
But not so much actually considering whether the
government is using the robots to kill people.
You know, there are robots that are designed to take to
kill people.
But they call it increasingly lethality.
But if you keep a little cheat sheet, you can figure out
what's going on in those contracts.
But I don't really talk about them that much in this book.
Because in my opinion, if you're standing in front of
giant robot with tank treads and a gun aimed at your head
and every time you walk like the gun kind of tracks like
that, it's pretty clear what kind of a situation you're in.
[LAUGHTER]
DANIEL H. WILSON: But if your Roomba is like doing doughnuts
and there's a red LED blinking on it and it's covered in cat
fur and you can't find your cat or something.
I don't know.
This is just sort of stream of consciousness at this point.
That's a little more sinister, because you live with it.
And it's in your house.
And then it's happening more and more.
We're getting more and more consumer robots.
So I tried to stick to that.
If that answers.
AUDIENCE: Have you played with Battlebots?
DANIEL H. WILSON: With Battlebots?
Yeah, I have a little plastic one and you hit the button and
it kind of goes like that.
But in general, I've always thought about BattleBots a
little bit.
At CMU, I don't know.
Mostly everybody seems to frown on them because they're
not autonomous.
BattleBots are these robots that fight each other.
And they're not autonomous for a reason.
And that's something I point out a lot in the book, is that
if you're going to have a robot that's making decisions,
it has to have a lot of sensors.
And sensors are really vulnerable.
So if you're going to have a BattleBot fight with two
autonomous robots, it would be like, they get out there and
they start cooking and one would hit the other one and
the camera would break and it'd be like, ah.
And then the other one would like try to chase it down
until it hit the wall and its camera broke.
And then they would be like this for
the next five minutes.
It's very boring.
[LAUGHTER]
DANIEL H. WILSON: But you don't have to jump straight to
full autonomy.

You always see people driving their BattleBots, and it's
usually like a guy and his eight-year-old daughter.
I've never understood that.
But anyway.
So the daughter is like, kill, kill, and doing this.
And the thing is like swinging it's jack hammer,
whatever it's got.
But a human is trying to aim it, and it's a
non-intuitive interface.
You're trying to control this remote control car type thing.
What you really should have is a sliding level of autonomy.
You need like some sort of sensor that says, oh, he's in
front of me.
Pow.
I haven't seen anybody do that yet, and so I'm waiting.
As soon as I have time I'm going to break the
whole thing wide open.
It's going to be amazing.
You might want to sponsor my bot.
AUDIENCE: So you could have a bunch of ones with spikes hang
around and observe while the big one goes in and fights.
DANIEL H. WILSON: Yeah, yeah.
There's some DARPA grants I've seen that talk about that.
Well, and the other thing is, it's more interesting actually
when they cooperate sometimes.
So there's like RoboCup where they have robots that play
soccer against each other.
And this is fun, because they have a lot
of different leagues.
They've got the--
where's Pat Riley?
They've got the really boring simulation league where robot
agents play on a screen.
That was a personal insult directed to Pat.
And then they've got like real robot leagues where they may
be this big.
And then they've got humanoids I think now.
And they've got AIBOs.
But the best of are the trash can-sized ones.
Because these babies can actually hurt people.
And they've got like human referees that are out there.
And they don't really know what the robots
are going to do.
[LAUGHTER]
DANIEL H. WILSON: You can't really, you know.
They're trying to referee and it's like a bunch of blind
eight-year-olds running around and whatever.
Sorry.
I think that about wraps up my answer to that question.
AUDIENCE: In a follow-up to the BattleBots question, do
you have any recommendations for humans to survive in a
post-uprising world where robots force humans
to fight for sport?
[LAUGHTER]
DANIEL H. WILSON: Man, you've got a future in Hollywood.
[LAUGHTER]
DANIEL H. WILSON: That sounds good.
Yeah.
I think that you're much less likely in the near future to
be forced to engage in blood sport with
fellow humans by robots.
You're probably much more likely to have your work
schedule optimized by robots.
And like our friend pointed out already, in that respect I
think we may have already lost, so, yeah.
AUDIENCE: When the future is realized and the robots put us
all in game parks, which will be the game park to angle for?
To be put into?
DANIEL H. WILSON: Yeah.
It all depends on what floats your boat.
I guess.
The mating park, I think, is the one to go for.
AUDIENCE: Somewhere down in the Caribbean.
DANIEL H. WILSON: Go for the equator.
You know, it all depends on whether we nuke the surface of
the earth to block the solar rays that fuel the power cells
of the robots.
Sorry, that's the matrix approach.
And don't say I condone it.
I don't like it.
OK, sorry.
Trailing off again.
Is that all?
OK.
Thanks.
[APPLAUSE]