This is kinda wild. "Will a robot take my job?" is one of the most
googled questions. It kinda makes sense. You know, when phones
were first invented, there was probably
horse messengers freakin' out all over the joint. Vinyl's making a comeback, but fidelity aside, why bother, when you can have
a curated playlist with thousands of songs right on your phone? Come to think of it, I can't tell you
the last time I shot a film... on film. See, change makes people panic,
but things evolve regardless, so we accept it, we move on. And now that we're entering the so-called fourth
industrial revolution, yeah, automation's
gonna replace some jobs, but it will also
create new ones, and probably give rise
to industries that never existed. That's the topic some of these
next folks are exploring, how machine learning is making things
more productive, more efficient, safer,
and cleaner. [director] And cut! That felt okay.
You wanna go again? -No, we're set.
-Okay. [director] Maybe just
step back on that mark. Here? -Yeah.
-Okay. [director] That's good. [sighs wearily] Right, yeah, I think I get it. [horn honks] [Downey]<i>
The trucking industry</i> <i> is at a crossroads.</i> <i> E-commerce is driving
demand for shipping</i> <i> higher than ever,</i> <i> but the harsh realities
of long-haul trucking</i> <i> are sending
professional drivers
looking for other work,</i> <i> leaving a 50,000-driver
shortage in the U.S. alone.</i> [Dr. Ayanna Howard]<i> There are
of course some industries</i> that will be impacted by A.I. in terms of labor workforce
being diminished, <i> but there are some industries</i> <i> that don't have
the workforce to sustain it,</i> <i> and so that disconnect,</i> that's where A.I. can fit in
very nicely. [Fitzgerald]<i> My name
is Maureen Fitzgerald.</i> I've been driving trucks
for 32 years, <i> and I was ready
to make a change.</i> Three, two, one. -[device beeps]
-[man] Engaged. [automated voice]<i>
Autonomous driving started.</i> [Downey]<i>
Approaching retirement,</i> <i>Maureen saw automation coming,</i> <i> and assumed
it would replace her.</i> <i> She was partly right.</i> <i> Automation is coming,</i> <i> but her job won't be lost.</i> <i> It will evolve.</i> I've been in this career
a long time, I've seen how everything
has been changing. <i>The monotony of coast to coast,</i> <i> and the pay being low,</i> <i> and, you know,
you're sleeping in a sleeper,</i> <i> you're sleeping
at a truck stop.</i> It's hard. It's a hard life. <i>I saw an ad for a test driver,</i> and I said, "That has to be
the perfect job." [alerts beep] [brakes screech] [Downey]<i> Okay, but this isn't
just about Maureen.</i> <i> We've all heard about
self-driving cars,</i> <i> but self-driving trucks,</i> <i> why are they setting the pace</i> <i> of the autonomous
driving industry?</i> [Pedro Domingos]<i> We already
have robotic airplanes.</i> <i>Most airliners fly themselves,</i> but self-driving cars
are a harder problem <i> because the roads have
a lot more things going on</i> than the air does, <i> and driving on the freeway</i> <i> is much easier
than driving in the city.</i> So we will see fleets
of automated trucks, I think, long before we see
self-driving cars in the city. The company of TuSimple
starts with a dream... <i> a dream that we wanted
to build the A.I. technology</i> <i> for autonomous trucks.</i> [engine starts] [Arda Kurt] Can you
give me one full rotation
to the right? The truck was originally
not made to be automated. Okay, can you give me
50% throttle? <i> We add our intelligence,
the servers,</i> <i> and all the sensors
to make this work.</i> We're in the sleeper
of a TuSimple tractor. <i> We have removed what normally
would be beds</i> <i> and refrigerators,</i> <i> and we've added our computer,
which is the brains of the A.I.</i> Driving the car
is what is called an "A.I. complete problem." <i> If you solve it,</i> <i> you solve every other
problem in A.I.</i> However, to solve it, you need to solve
every problem in A.I. <i> So it requires vision,</i> <i> it requires robotic control,</i> so it requires
motion and navigation, <i> but it also requires
social interaction</i> <i> with the other drivers,</i> <i> so at the end of the day,</i> <i> in order to drive a car well
in the city,</i> <i> you need everything.</i> It requires an enormous amount
of common sense. [Downey]<i> Common sense.</i> <i> That's what Maureen
is doing here...</i> <i> overseeing the A.I. truck</i> <i> as it learns to navigate
real-world conditions,</i> <i> ready to take over
if anything goes wrong.</i> <i> Today, they're taking
a ten-mile loop</i> <i> to test scenarios that
a human driver might encounter.</i> <i> Can this 10-ton vehicle</i> <i> safely navigate
a busy freeway?</i> <i>Can it merge and change lanes?</i> [man] All right, ready. Three, two, one. Engaged. [automated voice]<i>
Autonomous driving started.</i> My name is David Ruggiero, and I'm a test engineer
with TuSimple. <i> My job is to let her know
what's happening,</i> <i> let her be
absolutely comfortable</i> <i> the truck is doing
what it's supposed to do.</i> Gear change...
There we are. <i> When the vehicle is engaged,</i> <i> she is fully aware</i> <i> of what's going on
around the truck,</i> <i> but she doesn't
have to touch anything.</i> [Ruggiero]
All right, clear on the right. [Fitzgerald]
Not clear on the left. [Ruggiero]
I see the left approaching. Never is one drive
or one day ever the same. [Fitzgerald]
The left is now clear. Is it gonna go? -We're good!
-Come on. [Fitzgerald]<i>
The truck was cautious...</i> So I'm still seeing
full braking. [Fitzgerald]<i> It was probably
overly cautious.</i> Just go manual? Yeah, we can do that. [automated voice]<i>
Autonomous driving off.</i> [Ruggiero] There we go. [Fitzgerald] All righty. -[device beeps]
-Engaged. [automated voice]<i>
Autonomous driving started.</i> Going 45. Looks good. Center lane. [Fitzgerald]<i> I feel like</i> <i> I have a relationship
with the truck.</i> Easy, easy... <i> I talk to the truck.</i> I tell it when it does good. <i> I've never been
a test driver before,</i> <i> and it's never been a truck
driving by itself before.</i> <i> We're both actually
learning together.</i> Still green. [Ruggiero] Still green... We're off throttle,
slight brake. Okay, red light. [automated voice]<i> Left turn.</i> [Fitzgerald]
Turn signal is on. [Ruggiero] I see the lead van. Human drivers cause 30,000
or more deaths per year. [tires squealing] Long-haul trucking
can be dangerous, it can be boring, <i> and therefore,
it's ripe for automation.</i> [Fitzgerald] On the right. [Ruggiero] Okay, I see
the other car moving, I do see green. Still green... Still green. [Ruggiero] I see no oncoming. We are making the turn. [Fitzgerald] That'll work. All right, the new merge ramp. [Downey]<i>
Merging onto a highway</i> <i> is difficult and dangerous.</i> <i> It requires mastering</i> <i> a complicated set of physical
and mental skills,</i> <i> having keen sensory
and spatial awareness,</i> <i>preparing for unpredictability
and human fallibility...</i> [Ruggiero] And we're in it. [Downey]<i> ...and doing it all
fast and fresh each time,</i> <i> otherwise you risk crashing,</i> <i> or worse.</i> Turn signal is on. [Ruggiero] Okay,
I'm showing a left intention. [Fitzgerald] We are gonna have
a truck next to us, but he's moving over. Okay. Confirmed. [Fitzgerald] Looks like we
are clear to get over, the pickup is moving. [Ruggiero]<i>
This truck is measuring</i> <i> what's coming up behind it,</i> <i> how fast they're
coming up behind...</i> Looks a little
congested up there. Yep. [Ruggiero]<i> It has
a full 360 view, all the time.</i> [Price]<i> The A.I.
uses the images</i> <i> that are coming
from the cameras</i> with other sensors, LiDAR, and radar. <i> LiDAR is a laser range finder</i> <i> that is measuring
the distance to objects</i> <i> 360 degrees around.</i> <i>It gives us a three-dimensional
picture of the world.</i> In the virtual world, <i> there is a direct
correspondence of the map</i> <i> and the location
where the vehicle is,</i> and the behaviors
of other vehicles, their locations, their speed, <i> their future intentions,</i> <i> and once all these things
are re-created,</i> it's basically asking
the computer to play a game. Okay, tire front,
continuing at 57. Good job. [Ruggiero]<i>
When I'm looking at the data</i> <i> that the truck is feeding me,</i> <i> not only am I seeing
the current environment,</i> but I'm seeing ahead
into the future. And we have a slow vehicle
cutting us off. I see the cut in, at a 51. [Fitzgerald] That was
a good one. [Ruggiero] Okay,
I saw the cut in. <i> If we need to make
a lane change,</i> the truck is going to know
before I do. Okay, I currently
don't have a-- There we go,
I have an intention. We're going right now. [automated voice]<i>
Left change.</i> [Fitzgerald] That's fine.
We're good. All right, let's go. [Downey]<i> Okay, so this is
where things get interesting.</i> <i> This truck's tricked out
with sophisticated cameras</i> <i> and software,</i> <i> but it still doesn't have
one critical element...</i> <i> the thing that A.I.
may never have.</i> The central problem in A.I. is that human beings
have common sense <i> and computers do not.</i> <i> We take common sense
for granted.</i> We know how the world works, <i> and everything that we do
in our daily life</i> <i> involves common sense...</i> Pretty steady stream now
on our left. [Ruggiero] I see it. ...and people have been trying
to imbue A.I. with common sense since the beginning, and what is extraordinary is, at the end of the day, how little progress we've made. [Downey]<i> The endgame
is that, one day,</i> <i> the AI truck will not need
Maureen, David,</i> <i>or any person to ride with it.</i> <i> Each one of these test runs</i> <i> is a building block
toward that goal,</i> <i> designed to help it learn
how humans operate</i> <i> and make decisions,</i> <i> even when you get thrown
a curveball.</i> [Fitzgerald] This truck
is messing with us. [Ruggiero]
Just catching the vehicle. He's all over the... -He's coming close to the line.
-Mmm-hmm. [Ruggiero]
I have a left intention. Uh, lane is blocked. We were having
a little bit of challenges from one of your camera trucks. [Ruggiero] All I need is a gap. Just looking for the gap. Come on, you can pass him,
it's okay. [Fitzgerald]<i>
The truck was cautious,</i> because it couldn't predict <i> what that camera truck
was doing.</i> Why is he going so slow? [Downey]<i>
The film crew's vehicle</i> <i> was shooting the truck
for the whole afternoon,</i> <i> and, yeah,
it wasn't driving normally.</i> The camera truck was
what we would consider a non-compliant vehicle. <i> It was doing things
that a normal driver,</i> <i>you wouldn't expect them to do.</i> [Fitzgerald] Doesn't trust it. [Ruggiero]
Okay, I see the cut in. Sixty and slowing. Fifty-two and slowing. Still seeing
a left intention. [Fitzgerald] Let's see.
Okay, we are no longer clear. No, it's full. [automated voice]<i>
Pre-left change.</i> We're not going. He's all over the place,
that guy. [Downey]<i> The truck wants
to pass the camera crew,</i> <i> but the constantly
changing speeds</i> <i> and unusual behavior</i> <i> confuses the AI...</i> <i> so it refuses
to take the risk.</i> [automated voice]<i>
Canceled.</i> There are behaviors
that the truck has to learn to become human, in some sense, <i> because not everybody else
is operating at the same level.</i> All right. Turns out that getting
the first basic systems was a big milestone, <i>but then there's this long tail
of difficult situations</i> <i>that are much harder to solve.</i> [horn blaring] <i> When you're merging,</i> <i> do you have to give way,</i> or does it make sense
to go forward? That requires
a level of understanding <i> that we humans barely have,</i> <i>but machines aren't there yet.</i> [Ruggiero]
That's definitely gotta be the most interaction I've seen
with one vehicle. [Fitzgerald chuckles] [Ruggiero]<i> The big takeaway
on today's run...</i> there is no such thing
as a perfect run. <i> Eventually,</i> <i> we hope that the trucks react
to a change in environment</i> <i> as quickly as we can,</i> regardless of
what we throw at it. [Fitzgerald]
Okay, slow down for the gate, and looks like
they're going to let us turn. [Ruggiero] Beautiful. [Fitzgerald] Thank you, people. We still have
a lot of work to do to validate the system, <i> assure that everything
is perfected,</i> and this will take
substantial driving experience, simulation experience. [Fitzgerald]<i> Watching
something develop over time,</i> <i> you never thought
it would learn like this,</i> <i> and now it is learning,</i> and you feel like
a proud parent. It was making
some awesome decisions that I've never seen it
make before. <i> I get asked
all the time by people,</i> you know, "Oh, you're taking
truck drivers' jobs away by having autonomous trucks," <i> and I don't believe
that's true,</i> because it's not
taking away jobs, it's taking the place
of a job that a driver
doesn't want to do. [automated voice]<i>
Thanks for riding
with TuSimple.</i> Ta-da! High five. [automated voice]<i>
Autonomous driving off.</i> [Fitzgerald chuckling] [Downey]<i> What Maureen thought</i> <i> was the twilight
of her working life</i> <i> actually became the dawn
of something new,</i> <i>and it's an age-old phenomenon.</i> <i> Tractors didn't make
farming obsolete,</i> <i> and video didn't kill
the radio star.</i> <i> It just provided
new opportunities.</i> <i> New technology is changing
how we think about work.</i> [man]<i> Hey, Garrett,
that east on 424
should be good.</i> 10-4. [Howard]<i> I think the biggest
changes that we face</i> <i> when we think about A.I.
and the future of A.I.</i> is what do we do when A.I. changes
what the job functions are, <i> and what the workforce
looks like,</i> <i> which it's going to do.</i> <i> So this is a fear,</i> <i> because change,
people don't like,</i> <i> and change that
you don't understand...</i> is terrifying
for a lot of people. [Anthony Otto]<i>
Port of Long Beach</i> <i> is the second-busiest port
in North America...</i> and the first fully-automated
container terminal in the United States. [Downey]<i> To most people,</i> <i> shipping and logistics
aren't that sexy,</i> <i> but to an A.I., it's a dream.</i> <i> With over 14,000 people</i> <i>moving more than eight million
giant containers</i> <i> around the world,</i> <i> this port handles about
$200 billion worth of cargo</i> <i> a year.</i> <i> Automating this industry</i> <i> will make it
more efficient, for sure,</i> <i> but also safer.</i> [Otto]<i> As you can see
this fence here,</i> <i> from this point on,</i> <i> all the way for about
an eighth of a mile that way</i> <i> is a fenced-off no-man zone,</i> running entirely
off of software. <i> In the old design,</i> <i> you had the vessel's activity</i> <i>mixing with the truck activity,</i> <i> competing for space,
congestion, safety issues...</i> <i> ...but the vessel sizes
just continued to grow,</i> seven times larger <i>than the ships that we ever had
in the beginning,</i> so there was a need
for a new model, which is what LBCT is. [Josh Johnson]<i>
When a ship pulls alongside,</i> we're using
different algorithms to figure out traffic,
scheduling, dispatching, and planning. <i> We're doing this</i> <i> with 10 different cranes
at the same time,</i> <i> 50 different vehicles
at the same time,</i> <i> and we have
30,000 different places</i> <i> to put that container
in the yard.</i> When we offload
a container from a ship, the first thing it's gonna do is get picked up
by a ship-to-shore crane <i> that does have
an operator inside.</i> <i> -Hey, Lily, it's Boyle.</i>
-[Lily]<i> Yeah, I copy.</i> [Boyle]<i> Okay, Lily,
which side you wanna start on?</i> [Johnson]<i>
Ship-to-shore cranes</i> <i> will set the containers
on the platform.</i> <i> On that platform,
there are over 20 cameras</i> that use
Optical Character Recognition, <i> or OCR.</i> <i> Another crane will then
pick it up,</i> <i> put it onto
a fully-automated vehicle.</i> <i> That vehicle will drive
through the yard,</i> <i> and it'll get to
the right spot,</i> <i> planned by the system.</i> What looks like
a little wad of bubble gum, these are transponders
buried in the ground. <i> They're in a grid pattern,</i> <i> there's more than 10,000
out on the production field.</i> <i> There's an antenna
on the front, and the rear,</i> <i> and they read them
as they drive,</i> and then transmit that
to the system <i> to let it know where it is.</i> Once it reaches the yard, an automated stacking crane
will pick it up <i> and deposit it
into a yard block.</i> <i> From the block,</i> <i> it'll eventually
get deposited onto a truck</i> <i> or onto a bomb cart,</i> where it can make its way
onto a train. I don't have my key. [man]<i> On my way.</i> That's why we work in teams. -[man]<i> What'd you say?</i>
-[laughs] [Downey]<i> They say
that we can eventually learn</i> <i> to adapt to an environment</i> <i> that mixes robotic
and human elements,</i> <i> but it's more difficult
for machines to adapt to us,</i> <i> so here,
they need to be separated,</i> <i> because of safety.</i> Because humans
are unpredictable, automation can be
physically dangerous to people, <i> because if I'm
a 200-ton machine</i> <i> and I don't see you?</i> I might just run you over. [winches clanking] [man] This is
the high-danger area. When we go out in the field,
we call for a restriction, and they block a certain area
of the production field, <i> and no AGVs know
to go in that area,</i> <i> and we can safely work.</i> You want these two right here? -[man] Yeah...
-Okay, cool. This facility,
it is fully automated, but it's not as if someone
comes in, pushes a button, and puts their feet up
on their desk. <i> It is still employing
hundreds of individuals</i> <i> on any given day...</i> How does everything look
on that side, Eric? [Eric] Got a broken rail
over here, Julio. We're gonna have
to weld it up real quick. Okay. Grab my hood from
inside the truck, please? Here we go, guys.
Watch your eyes. Eyeballs! You know, this is
an asset-intensive facility. <i> The number of mechanics
necessary</i> <i> to maintain this facility</i> <i> has certainly grown.</i> [Julio]<i> My background
is automation.</i> That's why I decided
to come here, because it was high-tech. [Johnson]<i> Instead of
a yard crane operator</i> in a cab alone all day, <i> now we're working remotely
in the operations room.</i> About 80% of the time, most of the faults
are resettable from here. [woman] Okay, easy fix. Easy fix. <i> I'm Garrett Garedo,
crane mechanic here at LBCT.</i> <i> This is the central
operations system.</i> <i> A bypass key
will allow the operator</i> <i> to bypass a function.</i> [Downey]<i> Automation
can make the workplace safer,</i> <i> and this is a good example...</i> <i>taking someone out of the yard,</i> <i> and putting them
into the control room.</i> Right here, I'm gonna need
to move this container over, because it's not
sitting too good. You ready?
-Yep, go ahead. [Howard]<i> A.I.,</i> <i> like any technology
that comes into our society,</i> <i> it changes the workforce.</i> <i> There's this misconception</i> that there will be fewer
of these new jobs. That's actually not the case. Let me move it back. [Howard]<i> There are new jobs,</i> <i> there are different
types of jobs.</i> [man]<i> I was a crane operator.</i> <i> Back in the day,</i> <i> we'd have to climb up
five stories into the cab.</i> It was a little hard,
especially on the back, <i> and then
when automation started,</i> <i> they just shifted me over,</i> <i> and now it's a lot easier.</i> [laughs] If you don't embrace it,
you're just gonna fight it, and it's not gonna... You gotta use it as a tool. Right on the money. [Otto]<i>
You know, our obligation</i> <i> was to re-train
those same individuals</i> <i> into the new jobs that
that technology has created.</i> We are the progress
that was needed for the goods movement industry. For the foreseeable future, there will continue to be
a lot of things <i>that only humans can really do.</i> So I've dispatched
a mechanic out to STS-1. Okay. [Domingos]<i> And so
the future of work</i> <i> is not humans being replaced
by machines.</i> It's humans figuring out
how to do their job better with the help of machines. [Downey]<i>
Structured environments</i> <i> are good for automation.</i> <i>Smooth surfaces, right angles,
less chance for chaos.</i> <i> But what about more complex,
human environments?</i> <i> Can we learn to work
with robots hand-in-hand?</i> [man] Oh, there we go. Wait a second. Stop. Let's see if that works. [Downey]<i> RoboHub is
a premiere robotics incubator</i> <i> at the University of Waterloo
in Ontario, Canada.</i> Gripper's not working
on the left. Maybe if I turn
a little bit... [Downey]<i>
One thing they're doing</i> <i>is developing A.I. and robotics</i> <i> to use in environments
that are unstructured,</i> <i> more human.</i> <i> Like at home.</i> Calm down, calm down. [Brandon DeHart]<i> TALOS
is one of the most advanced
humanoids on the planet.</i> <i> It can walk, it can talk,</i> <i> it can see you in 3D...</i> but it can't do
most of those things out of the box, so you really have
to take it as a tool <i> and teach it how to do
a lot of these things.</i> On TALOS, we want to explore
two different aspects of A.I. <i> First, to perceive objects
in the world...</i> [DeHart]<i> So it has cameras
in its eyes,</i> <i>and it also has a depth camera</i> <i> where it can actually see
how far away things are</i> <i> within its vision,</i> much like we do
with our depth perception, <i> and as soon as TALOS
has that ability to see,</i> <i> "Oh, there's a human here,
there's an object back here,"</i> then it can start building
a map of its world, so that it knows in 3D
what is all around it. [Howard]<i> Computer vision</i> <i> is a way to mimic
how we see the world.</i> I can say, "Well, that's a TV," or "That's a cat,"
or "That's a dog," so what is it
that I'm looking at when I'm looking at an image? <i> Computer vision is trying
to figure out</i> <i> what are the objects
that are in this image</i> that represents
what we would understand. And now I can teach the robot, basically by grabbing it, and I will just gently
move the robot around, and guide it
towards the bottle... and you just hit
the right trigger button to execute the same motion. Should I close it? No, no, it will do that
automatically this time, because I already kind of
integrated this. [Werner]<i>
Also we want to explore</i> application of A.I. <i> in understanding locomotion</i> <i> in terms of balancing.</i> No, it's going to crash. [laughs] Played too much. The biggest misconception
about robots is that they are
more capable than they are, that they're more generalized
than they are... [clattering] ...and they're not quite
there yet. [Werner] Okay, let's give it
one more shot. [DeHart]<i> So you may have seen
humanoid robots</i> <i> that can do parkour,</i> <i> do backflips.</i> The only ones
that are really comparable are the Valkyrie
that was designed by NASA, <i> of which there's only
two or three in the world...</i> <i> and the Atlas
from Boston Dynamics,</i> but most of them
have very few sensors, <i> and a lot of the time,</i> <i> they will be
remote-controlled,</i> <i> or given a very clean,
pre-scripted path,</i> and in a situation like that,
what they're doing <i> is pushing the limits
of the mechanical systems.</i> That's very similar
to what we're doing here, except on our case, we're
pushing the software side. [Brynjolfsson]<i>
The next step going forward</i> is to start replacing
all of our dumb, blind, <i> dangerous machines</i> with machines that have sensors
and vision built into them, that can work
side by side with humans to be more productive and safer at the same time. [Werner] Maybe we can
carry a table
with the robot together? So if we put the table
in its hands... [DeHart]<i> So for a task</i> <i> where the robot might be
carrying something
with a human,</i> <i> it's going to need to be able</i> <i> to feel how hard the human
is pushing, pulling,</i> much like if you're
being guided in a dance. Can we also
tighten the grippers, Alex? Yeah, yeah. So as humans,
we have a sense of touch, and we have a sense of
how much force we're applying, and is being applied to us. <i> With TALOS,
it doesn't have a skin,</i> <i> it's hard plastic everywhere,</i> <i> so it can only really sense
what's in its motors...</i> [Werner]
This will take a second. We have to teach it
how to translate that into sort of a sense of touch. Try to keep it
somewhat level, but if you kind of
twist the table as well? [Werner] Let me show you
how much is actually possible. The one who writes
the controller is always braver
than the rest of us. [woman laughs] [Werner] And you just
start pulling a bit... and then the robot
will walk with you. <i> TALOS is using
compliant control</i> <i> in the upper body,</i> which basically
makes him soft. <i>The robot automatically senses</i> <i> the forces being applied
to the table.</i> The human is still
fully in control. <i> This is very important</i> <i> to the safety of humans
surrounding the robot.</i> Within robotics, the areas where you're really
gonna see the important advances <i> are those environments</i> <i> that are relatively
controlled and predictable,</i> <i> so a good example
is an Amazon warehouse.</i> In those environments,
you already see lots of people, <i> and also lots of robots...</i> and they're working together. [Brynjolfsson]<i>
Once you have a process,</i> and you've reduced it
to an algorithm, <i> you can replicate it</i> <i> so that if a robot can learn
a whole new skill,</i> <i> you can copy that knowledge
through the cloud</i> <i> to all the other robots,</i> and now they all have
that same skill. <i> It's a whole new world,
a whole new kind of economics,</i> and we're just beginning
to understand its implications. [horn honking] [Downey]<i> A whole new
automated world</i> <i> still hard to imagine?</i> <i> Maybe it's more about
what we're gaining</i> <i> than what we're losing.</i> <i> Will automation
make things faster?</i> <i> More efficient?</i> <i> More productive?</i> <i> Tastier?</i> [man]<i> I've been making pizza
for 15 years.</i> In my life, I've probably made
two million pizzas... so coming to Zume was awesome, because it was like, wow, this is different
than any other pizzeria. [Alex Garden]<i>
So I got the idea for Zume</i> <i> about seven years ago</i> when I realized that Big Pizza was built
on one central conceit... <i> which was that there wasn't
a better way to do it.</i> I thought, "I wonder if there's
a better way to do it?" [Downey]<i> Big Pizza.</i> <i> Doesn't have
the same ring to it</i> <i> as Big Data, or Big Pharma,</i> <i>but it's not just about pizza.</i> <i> It's also about big waste.</i> <i> Last year in the U.S.,</i> <i> about $200 billion
worth of food was wasted.</i> <i> $200 billion.</i> <i> That's a lot down the drain,</i> <i> and also up in the sky.</i> <i> Waste is a huge contributor
to greenhouse gas emissions.</i> [thunderclap] Perhaps half of all food
that's produced in the world <i> is wasted.</i> Artificial intelligence can
help us change the supply chains so that they are
much more efficient. <i> If we can then
have an impact there,</i> that already
has huge, huge implications. One of the classic tensions
in every food business <i> is producing too much,
and having waste,</i> <i> or not producing enough food,
and having stock outs.</i> <i> It takes years and years
of practice</i> <i> for a human to forecast that,</i> but that's actually
a data problem that machines,
and A.I. in particular, are really, really good
at getting right. <i> So at its core,</i> Zume is a fundamentally
new logistics model. <i> We predict
what we're going to sell</i> <i> before you even order it,</i> <i>and that whole predictive layer
of the business</i> is all driven by A.I. [Chris Satchell]<i>
When Zume first called me,</i> <i> I obviously thought,</i> "Why on Earth
does a pizza company want a Chief
Technology Officer? <i> But then
when I heard the vision,</i> <i> the fact that</i> <i> we were going to be able
to change an industry</i> <i> by focusing on this amazing
end-to-end platform,</i> which just happens
to also produce amazing pizza, <i> then it made sense.</i> [Downey]<i> To disrupt Domino's,</i> <i> Zume uses machine learning</i> <i> to try and forecast
how much pizza people want,</i> <i> what kind, and when.</i> <i> It crunches dozens
of different variables...</i> <i> location,
day of week, weather,</i> <i> what's on TV,
past ordering trends,</i> <i> and then predicts
how many Sgt. Pepperonis</i> <i> San Francisco will want
on a Tuesday,</i> <i> for example.</i> Our supply chains
are incredibly inefficient. Our food processing systems
waste a lot of food. <i> If we could improve
predictions,</i> <i> we could eliminate
most of that waste.</i> We'd know what the demand
was going to be, <i> where the products
were going to be,</i> and ultimately, that would
make us all better off. Absolute perfection would be
no pizza gets wasted, and anytime a customer looks, there's always one of the type
of pizzas they would like available. So it's an optimization problem. How close can you get? [Downey]<i> Optimization, right,</i> <i> but what else?</i> <i> "End to end"
is not just about prediction.</i> <i> It also involves automation,</i> <i> or...</i> <i> cooking.</i> Pizza and robots,
very exciting concept. Pepe and Giorgio,
they dispense the sauce. We have Marta here, who spreads the sauce
on the pizzas... So we're still on
Veggie Zupremes, guys. These are looking good. [man] ...and down here
at the end, we have Bruno, who helps deliver the pizza from the assembly line
to the ovens, or around the bypass, and in the back, Vincenzo, who helps load the pizzas
onto the cartridges. [Downey]<i>
Based on its algorithm,</i> <i> the A.I. predicts
how many pizzas, and what kind,</i> <i> to load
onto the mobile kitchens.</i> <i> When orders come in,</i> <i> it also decides which
mobile kitchen will bake it,</i> <i> and exactly when
to put each one in the oven.</i> Every single pizza has its own cooking profile
and recipe. So each one of these ovens
actually is a robot. They're connected
to cloud services <i> that're always monitored,</i> making it possible
for one person working in here to cook up to
120 pizzas an hour. [Zaida Ybarra]<i> It's humans
and technology
working together.</i> <i> It's pretty awesome.</i> That and the pizza. I love pizza. Pizza is one of
my favorite foods. We use A.I. in simulation around how do we get <i> the right delivery
estimated time of arrival?</i> <i> The A.I. can suggest</i> <i> here's the best place
to put a truck</i> <i> so that we get the hottest,
freshest product</i> <i> to our customers.</i> So we have
a BBQ Chicken. BBQ... [Ybarra]<i>
When we get the orders,</i> <i>our system controls the ovens,</i> which controls the cook time, <i> and how long the pizzas
need to cook.</i> [Satchell]<i> The A.I.
can change the workflow</i> on one of our vehicles to make sure the pizza's
cooked at the last moment. ...and then the system chooses
which driver to send it to. Here's your order... A BBQ Chicken ranch. It's algorithms plus,
it's A.I. plus humans, not A.I. instead of humans. Now we have a pineapple
that is ready to go. [Ybarra]<i> Their predictions</i> <i> for the amount of pizzas
that we carry on the truck</i> <i> are pretty spot on.</i> Last Zupreme of the night. Sometimes we'll have
a bunch of pizzas left over by the end of the night, <i>and we'll think the prediction
was probably wrong for today,</i> and then boom,
an order comes in, <i> and the next thing you know,
another order comes in,</i> <i> like, eight pizzas at once.</i> All of that data gets fed back
into the learning algorithm, so every week <i> we're trying to evolve
our learning algorithms</i> <i> to do a better job
than the week before.</i> <i> Part of what we're
optimizing for</i> <i> is reducing waste.</i> <i> Maybe if you can use
prediction</i> to go back
into the supply chain and predict more accurately
what you need to grow, <i> and where you need
to get that product,</i> I think you could fundamentally
change things. We're in the early stages of a massive change
in technology <i> that's allowing machines
to do a lot of tasks</i> <i> that previously
only humans could do.</i> [Downey]<i> The future of work
is evolving.</i> <i> Automation's making
the workplace safer,</i> <i> greener,</i> <i> more efficient, for sure.</i> If we can figure out how
to integrate A.I. technologies, <i> not to replace humans,</i> <i>but to augment their abilities,</i> <i> to make life more fun,</i> <i> to make us more productive
and more creative...</i> I think that's where the power
of the machine will come. [Downey]<i>
Thousands of years ago,</i> <i> we were hunter-gatherers.</i> <i>Eventually, we became farmers,</i> <i> and now that
we're maybe entering</i> <i> this fourth
industrial revolution,</i> <i>one that connects the physical,
biological, and digital,</i> <i> it's not just jobs that are
being changed by A.I.,</i> <i> but us.</i> <i> We are the progress.</i> <i> So why focus on the rear-view</i> <i> when we can look
to the road ahead</i> <i> to see what we're becoming?</i> [woman sighs] [Ford]<i> It's a real challenge
in any case</i> <i> to build robots</i> with the kind of mobility
and dexterity that begins to approach
what human beings have. <i> Kind of the cliche
that we all think of</i> <i> is that
we'd like to have a robot</i> that can go to the refrigerator
and grab a beer for us. I mean, if you think about
what's involved with that, that's still
an enormous challenge. <i>A robot that's able to do that</i> <i> has to be heavy enough
to go to a refrigerator.</i> <i> It has to be able
to open the door,</i> and it has to have
the visual perception and the dexterity to reach in
and grab the beer, and already that implies
a fairly heavy machine, <i> otherwise
it would just tip over.</i> <i> Building a robot to do that
is not just difficult,</i> <i> but it's gonna be
fairly expensive,</i> <i> so I think that
it's gonna be quite a while</i> <i> before we really see
the kinds of robots</i> that we imagine
from the science-fiction world operating in our daily lives. [robot]<i> Get ready
for a picture.</i> <i> Smile, please.</i> [camera shutter clicking]