April 25, 2024 • 37 mins
Imagine picking up your phone and ordering something from Walmart. Fifteen minutes later, a drone hovers over your yard, lowers your order down to you, and zips away. Adam Woodworth wants this to be so boring you don't even notice. He’s the CEO of Wing, a drone delivery company. His problem is this: How do you turn a flashy idea like a delivery drone into something as ubiquitous as a shopping cart?
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:15):
Pushkin.
Speaker 2 (00:20):
If you live in the Dallas Fort Worth area, there
is a good chance that right now you can open
up an app on your phone and order a cup
of coffee or a bottle of ibuprofen from a nearby Walmart,
and then about fifteen minutes later, a drone will appear
hovering in the sky above your yard. The drone will
lower down a rope that holds a box that has
(00:43):
your ibuprofen in it. Then the drone will leave the
box on the ground, pull up the rope and fly away.
People have been talking about drone delivery for a long
time now. There have been pilot programs and test sites,
but this thing in Dallas is big. Millions of people
live in the delivery zone. It's the first place in
(01:03):
the US where for lots and lots of people, drone
delivery seems set to become a normal, everyday thing. I'm
Jacob Goldstein and this is What's Your Problem, the show
where I talk to people who are trying to make
technological progress. My guest today is Adam Woodworth. Adam is
(01:26):
the CEO of Wing. Wing is owned by Alphabet and
it's one of the two drone companies doing deliveries for
Walmart in Dallas. Wing has also been making deliveries in
Australia for several years now. Adam's problem is this, how
do you turn a delivery drone into something as boring
and ubiquitous as a shopping cart? To start, I asked
(01:48):
Adam to make the case for his project, like, why
should I be happy to learn that delivery drones may
soon be coming to a sky near me delivering coffee
and burritos to my neighbors.
Speaker 3 (02:01):
I've caught a lot of flack over the years for
the burrito copter, and like, oh, you're just using drones
to fire around and deliver cups coffee and stuff. But
at the core of it, like people eat food every day,
Like if you look at it from a holistic transportation perspective,
like people are not going to stop ordering things, Like
(02:22):
that's not like it's in fact the opposite. People are
ordering more stuff every day. And it just does not
make sense that we are moving around like burritos and
cups of coffee in cars, Like it just does not
make sense.
Speaker 2 (02:38):
To what extent are people ordering coffee delivered by somebody
in a car? Is that a thing?
Speaker 3 (02:42):
And I just don't know it. It is a thing.
It is a thing like you can go sort of
look up. There's been a myriad of different studies about
sort of what actual sort of like dipler use cases are,
and it's a lot of small items. I can I
can talk about a bunch of aviation math, but I
think the simplest way to look at it is we
are designing systems like to be able to fly in
(03:04):
the national airspace. This is the same for everybody who's
sort of in here to be building systems that are
like sort of equivalent level of safety to like airliners.
So if you take that and you abstract it down
to things that people can sort of internalize in terms
(03:24):
of sort of risk, it is safer to fly the
package via drone than it is to walk the package
in your hand down the sidewalk.
Speaker 2 (03:36):
That's good, that's compelling, that's fun. Non intuitive, it's very
non intuitive. I thought you were gonna say safer than
to drive, and I'd be like, sure.
Speaker 3 (03:43):
Everything everything, but like safer than to hand carry it.
Speaker 2 (03:47):
The risk of hand caring is getting hit by a car.
Speaker 3 (03:50):
Yes, that's it's usually serve pedestrian road risk. And so
you have this form of transportation which is possible and
real and financially viable. That is many orders of magnitude
safer than the incumbent. If you look at on every
metric from safety to emissions, sort of impact on infrastructure,
(04:11):
It's like, it is silly that we are moving around
these boxes and the way we do today. This is
a fundamentally better way to do it.
Speaker 2 (04:18):
So if I want to order something, I live near
Walmart in Dallas and I want to order something from you, Like,
how does that part work?
Speaker 3 (04:25):
So basically like we built a pretty traditional looking marketplace
where the order flow is very traditional until you get
to the part at the end, which is the drone part.
Speaker 2 (04:37):
So I use your app. I don't use like the
Walmart app.
Speaker 3 (04:39):
So it depends on which region you're in. So in
Australia we're integrated into the door Dash app, So use
the front end of door Dash. In the US right
now we're integrate. We're using our own app. Okay, So
you say what you want ordered, and then before you
even get to that step, the back end will say
is your address deliverable? Right? Like it is the place
(05:01):
that you are, a place that the plane can get
to just to be.
Speaker 2 (05:04):
Clear that is you checking it out. You're using basically
satellite data to say does this address work? Or that's
you asking music. No, that's you just checking.
Speaker 3 (05:12):
So it's just basic. And some of it is even
like like if there's a big hill that the plane
has to spend more energy climbing over, like you might
be within the radius, but like your particular house like
may be undeliverable, or you might be slightly outside of
the radius because there's no there's no power lines or
hills to get there, so the plane go slightly farther.
(05:34):
So everybody thinks about a delivery area as a circle,
but it's really like a weird amiba.
Speaker 2 (05:41):
With lots of variables affecting where you can go.
Speaker 3 (05:43):
Yeah, and so there's a there's a sort of there's
an address checker that'll run that like just GPS coordinate
through a terrain database to be like can the plane
get there? And then once you get to check out,
you'll be presented with either one delivery zone of like okay,
this is the place on your property where the box
can be dropped off, or a choice like if there's
(06:04):
multiple spots, it'll be like okay, which one of these three,
do you want it in the left side of your
yards sell the ardor in the backyard the personal pick that.
Then the order will process, the plane will be dispatched,
and then it'll go to like a sort of countdown
of like, Okay, the plane is on its way and
it'll be here in three minutes and forty two seconds.
Speaker 2 (06:25):
So uh, and and what's the cost?
Speaker 3 (06:29):
It depends it's uh, it's sort of uh. That's that's
that's more of a partner question.
Speaker 2 (06:35):
What's it what's it cost you?
Speaker 3 (06:37):
That's a that's an unanswerable question.
Speaker 2 (06:40):
A lot right, now, A lot right?
Speaker 3 (06:43):
No, I mean I think, uh, I think less than
less than what folks would assume.
Speaker 2 (06:49):
I mean, the marginal cost is low, right, but it's
not that many deliveries against a large fixed costs.
Speaker 3 (06:54):
That's fair. But the marginal metrics are all sort of
you know, favorable or in the direction of favorable. Yeah. So,
like you know, people are well accustomed to getting stuff
delivered today, and so like those costs are well understood,
and the sort of costs per delivery that you know,
(07:14):
either is either the partner is charging they're sort of
all in line with that, depending on sort of which
region you're in. But that's that's ultimately up to the partner.
Speaker 2 (07:23):
A few bucks, I mean, is is that what that means?
Reason which you know it can't be more than that
if people are out in coffee, right though, it is
amazing how much people spend on, you know, on the
marginal thing. So so that's how it works from the
consumer's point of view, How does it work from you know,
from the drones point of view. From from the company's point.
Speaker 3 (07:42):
Of view, what happens every every morning the service turns on,
Folks can can order the stuff. Planes will get automatically
tasked to go take off, pick up the box and go.
Speaker 2 (07:54):
How many deliveries what they do in a day?
Speaker 3 (07:58):
Like our high water mark right now is around for
one store was about like one hundred and thirty deliveries, okay,
So it's like these are like, these are real numbers.
This isn't like like, oh okay, like three people ordered
stuff today, And I think that that's the that's I
think sort of the phase that the industry is moving
into now is it's it's these aren't like pilot programs
(08:21):
and trials. These are like, okay, this store opened up
offering this service and all the people in the area
expect to get that service.
Speaker 2 (08:28):
How how big and how heavy can something be?
Speaker 3 (08:31):
So you know, we carry about a little bit over
a kilo of of of good, so you know, two
point two pounds of stuff, and in.
Speaker 2 (08:39):
Terms of volume, like.
Speaker 3 (08:41):
Like I think about like a shoe box, shoe box
ish shoe box, okay, And a lot of it ends
up being you know, prepared food, drinks like you know,
soft drinks, you know, grocery items. And I think the
one of the interesting observations I've always had on this is, uh,
back back to the sort of original like did we
(09:03):
pick the right use case piece. We did a bunch
of surveys sort of early on in the program of
you know, what would you want delivered by drone if
you if you had the option, and you know, there
were dozens of different options sort of from food all
the way to medicine, clothing, all these other things, and
overwhelming the responses were always like, you know, drones should
(09:26):
only be used for medical delivery, like they should only
be used for medicine first aid bits like that. And
then we went and offered the service in Australia. We
had we had all those goods in there. We had
like a lot of you know, like type of you know,
over the counters, pharmacy items, sort of food, clothing, a
(09:48):
bunch of different local restaurants, and like eighty percent of
people ordered coffee. Like that was the overwhelming sort of
thing because it's like that's something that people got every day,
and it made.
Speaker 2 (09:59):
Sense stated preference versus revealed preference. I mean there is
there is also the externality thing, right, Like it's one
thing to say, if a drone's going to be flying
over my house all the time, what do I want
it to be delivering to everybody? Versus if it's coming
to my house, what am I going to order today?
Speaker 3 (10:16):
Right? I mean that is another way to frame that's fair.
Speaker 2 (10:19):
So like what what are when you look back after
ten years of working at it and like now you're
actually doing it, Like, what are a few key things
you figured out?
Speaker 3 (10:30):
I think one of the most defining moments is or
maybe one of the most defining constraints. Most sort of
disruptive or merging technologies sort of start off in a
permissive environment. So you know, one of the one of
the things that I you know, one of the questions
I've gotten asked many times sort of over the years,
(10:52):
is well, why didn't you take the approach that like
ride share took of. Okay, you know, offer the service
in a big metro area, build up a big customer base,
and then as the rules change, you have, like you
have customers to say, Okay, this is a valuable thing.
Speaker 2 (11:08):
Why didn't you do it?
Speaker 3 (11:10):
Why not just go? Why not why not just go?
Do it? And I think at the core of it,
it's because aviation does not operate that way right, Like
aviation is is is built around a preempted set of
rules where the federal government controls the airspace, so it's
not up to sort of anyone given municipality. And from
that place, you know, how do you build consumer demand,
(11:35):
How do you show that this thing works? How do
you even go and do a trial where the existing
regulatory frameworks are just say no, you can't fly at all.
Speaker 2 (11:44):
Yes to Australia.
Speaker 3 (11:47):
Yeah, So that's that's what That's what drove us to
to start in Australia because they were one of the
first markets where the National Aviation Regulator there they sort
of quantify the risk of the operation. They look at
the specifics the operation and they sort of approve on
a case by case basis of all right, like this
meets the same level of safety as an airliner, even
(12:09):
though it's a very different use case.
Speaker 2 (12:12):
So just tell me like one thing that you know
you had to figure out. Maybe you tried something and
it didn't work, and you tried something else and it worked.
You know what I mean? Is there just one kind
of a microcosm example.
Speaker 3 (12:24):
If I could do two a lot? If that's yeah.
So the first one is sort of how do you
achieve aerospace reliability with like consumer electronics costs? So you know,
we started from a place of how do you build
these planes in a way where the unit economics will work?
(12:45):
Not how do you build these planes and then try
to figure out how to make the unit economics work?
Speaker 2 (12:49):
Right?
Speaker 3 (12:50):
So what is I mean?
Speaker 2 (12:51):
Obviously the cost of one is going to be a
function partly or largely of scale. But since you brought
it up, like, what has one of them got a
cost for it to work for you?
Speaker 3 (13:00):
I mean it has to look like cell phones basically.
And I think that the interesting one that the sort
of specific point here is is if you looked at
just the reliability case Okay, so how many propulsion system
elements do you need in order to be able to
make the airplane sort of close all the reliability numbers.
(13:21):
That requires eight motors in our use case, so you know,
you need eight hover motors so that you can have
any one of them fail at any given time and
the airplane can still sort of have controlled flight. In
our cost trades, it ended up being that if you
move to twelve motors instead of eight motors, so the
(13:43):
reliability numbers close at eight. But if you went to twelve,
you got to use a much smaller motor that didn't
need a carbon fiber propeller. You could use a plastic propeller.
And there was a huge change in the cost of
the total propulsion system by increasing the number of motors,
which is not a thing that you would on the
surface look at and be like, okay, this is intuitively obvious.
Speaker 2 (14:08):
So you're using twelve cheaper motors, so more smaller. It's
more smaller motors rather than fewer larger motors. And I
mean presumably you could keep going that way. You could
have twenty tiny motors. I mean, you find the optimum.
Speaker 3 (14:25):
What settled on twelve was really driven by noise. So
like as you get the burders smaller, they have the
you know, the frequencies go higher, they spin faster, that
discon goes up all these different effects. And so moving
to twelve let you get the sort of cost benefits
by going small motors. But you didn't sort of start
(14:47):
climbing up that that noise curve by having like you know,
like one hundred little half inch motors hu and even
that there isn't an off the shelf propeller solution that
sounds pleasant.
Speaker 2 (15:02):
Yeah, So that the sound, the sound is a big one, right,
Like there was a good Wall Street Journal story I thought,
uh from Australia where you were doing a lot of
the early work, and the headline was delivery drones, cheer shoppers,
annoy neighbors, scare dogs. Tell me about that piece of it.
Tell me about the noise.
Speaker 3 (15:22):
So that's a that's a it's a thing you have
to be conscious of. So, you know, we did a
bunch of work early on in the development of this
aircraft around sort of how do we how do we
shape the tone of the noises being generated. So it's
sort of less of like one big peak and sort
of more broad spectrum, how do we reduce the noise
of it? Like, if you look closely at our propellers,
(15:44):
they're they're like asymmetrics. So there's one long set of
blades and one short set of blades, and that creates
like two tip forto sees that destructively interfere and like
spread out noise. So all these things. So we did
all this work on on the hover propulsion system, and
the vast majority of of sort of noise feedback from
the communities and mind you, like, if you compare this
(16:07):
to the noise feedback you get about an actual airport
or you know, construction project is still like far less
than that. But it was all overflight, so it was
like people weren't complaining about the airplanes hovering. They were
complaining about them flying over overhead. And if you you
(16:27):
look at the fact that the person that ordered the goods,
like the plane is coming, then they asked for it, right,
Like they're like they're like, I want the service. Their
neighbors are probably also ordering things. But the thousand homes
that you flew over to get to that spot, like
may not be participating in the service.
Speaker 2 (16:45):
It's a negative externality, right, you're imposing a cost on
someone who's not a party to the transaction.
Speaker 3 (16:51):
So so that was that was a very interesting learning
sort of early on in our that we only would
have learned by actually operating the service.
Speaker 2 (16:59):
Huh, it's that it's not it's not the next door
neighbor who hates it. It's all the people between the
store and the customer.
Speaker 3 (17:06):
And in that sense, it was like, okay, well, well,
like we thought we needed to put all this development
work in the hover propulsion system, but it's actually the
crew system that they need to work.
Speaker 2 (17:16):
People never get mad about what you think they're going
to get mad about, right again, I.
Speaker 3 (17:20):
Think that that's the at the core of it, Like
the dron delivery space is like so full of has
been so full of what ifs for so long, and
in the absence of regulatory frameworks that let you go
and operate like you can't move out these things that
it it oftentimes like there's a lot of misconceptions about
(17:41):
either noise or sort of what the use cases will
be or what people want to order, and then when
you go and offer the service, sort of the cure
for the what ifs is the actuality of an operating.
Speaker 2 (17:52):
So okay, so you realize that the noise complaints aren't
the noise complaints you're expecting. You had mitigated the one,
the hovering, but not the other, the flying over. So
now you have this new challenge, right that people getting
mad if a drones are flying over their house all
day long? What do you do about that?
Speaker 3 (18:08):
So we we basically did the same thing we did
on the other propulsion side, So we used the same
technology as to redesign the props. We sort of set
the design target so that like basically the aircraft that
the cruise altitude would be inaudible to a to an
observer on the ground. But this is like everything in
aerospace is a trade and so you know, for us,
(18:32):
it was like we want to we want to address
sort of any of the noise concerns. Uh, And so
we're willing to go and do that development time and
exercise and costs, redo the propulsions, and then redo the
plane to to be able to use it. So that
that's uh, that was that was one interesting one and
(18:54):
the other one is just like you know, how you
get the box to the people. So early on in
the program, uh, we had basically you know, you have
the little tope that hangs below the plane, and then
it was connected to playing with a tether, and that
tether went around a school and just sort of like
you'd open up a latch and it would descend and
(19:17):
you'd leave behind the string. So it was like the
way of letting go the box was just well, there's
only so much string on the on the on the aircraft.
Speaker 2 (19:26):
So so the customer would get a box with a
long string.
Speaker 3 (19:30):
Not a great user experience, but very simple, like super simple.
And uh, we spent a huge amount of time around
how do you drop off the box without a bunch
of mechanical complexity? So how do you how do you
lower down the box, make sure that it lets go
(19:50):
of the box one hundred percent of the time. Make
sure that that little hook at the bottom doesn't get
snagged on anything else as it's getting pulled back up. Uh.
Making sure that that whole system could operate reliably, you know,
thousands of times with dirt and leaves and stuff. Uh.
And so like this is little, that little yellow bit
(20:11):
of plastic. So you're just.
Speaker 2 (20:13):
Talking about the hook at the end of the line
that holds the box. Like that is like a nightmare
to get.
Speaker 3 (20:19):
Just right, nightmare to get just right. And it is
probably one of the most iterated components on the whole aircraft.
Speaker 2 (20:27):
Still to come on the show. Why despite all that
work on the hook and on everything else, drone delivery
may never really take off. Also, what the world will
look like if drone delivery does work? So what like
(20:53):
what's the frontier for you? Like, what are you trying
to figure out that you haven't figured out yet?
Speaker 3 (20:59):
Yeah, I think that for us right now. A couple
of years ago in Australia when we were sort of
ramping up volumes there, we demonstrated what scale looked like
for a sort of confined geographic area. So, you know,
one of the common I think misconceptions about drones is,
(21:19):
you know, folks like, well, when drone delivery is real,
I'll look up at the sky and all I see
is drones. And I was like, well, no, when you
when you actually do it, when you actually offer sort
of the scaled out service to like a you know,
one hundred square kilometer piece of the Earth, even when
there's even then when there's a thousand deliveries happening in
one day, and that that little space you can look
(21:41):
up and not see any planes for a while. And
so for the last couple of years that's what we
were focused on was Okay, how do we demonstrate the
unit scale of Okay, this is what drune delivery looks
like at scale for one one small space. Now I
think with the DFW expansion, we have the opportunity to
(22:02):
do that at a whole sort of metro wide scale.
So I think that the ability to offer the service
across a big metro to like, you know, where you're
gonna cover millions of people. Let's you answer, let's you
sort of validate that those assumptions that have been proven
out at smaller operational scales hold as you go up
(22:23):
and so looking at things like how frequently do people order,
like is it is it a novelty that people order
once and then they're like, okay, well I got my
cool TikTok video of a drown delivering me something or
do they find value in it? You know, do you
start to get to the place where you know, for us, like,
you know, is one operator able to you know, do
(22:47):
we have the permissions and is it working to have
one operator do multiple locations and you know, many aircraft
in the sky at once looking less and less like
sort of a one to one relationship between sort of
the person the operation to like more of a sort
of dispatcher view of the world of like, okay, this
person is managing a fleet of airplanes rather than energing
(23:09):
sort of anyone flight.
Speaker 2 (23:11):
So in the in the current service as it exists, Now,
what what is the role of human beings?
Speaker 3 (23:19):
It's it's really is like, uh, you know, a person
has to put the stuff in the box. A person
has to bring the box to where the plane is
going to be, and you know, the plane hovers and
lowers down the little hook and they hook the box
on it and it takes it up. And then you know,
somebody in the morning needs to take the planes and
put them out on the pads, and at night take
(23:41):
them and put them back in the shipping container. And
then our pilots are sort of monitoring the operation.
Speaker 2 (23:46):
I mean, just to be clear, they're not flying the planes, right.
The planes are flying themselves.
Speaker 3 (23:51):
Planes are the planes are flying themselves, but they're responsible
for the operation, right. So like if if you look
at the sort of like some of the core tenets
of aviation, like the pilot is responsible for the airliner
flight or for the charter flight. Like they're responsible for
the operation. And so what does that mean when things
get very highly automated, it's okay, well they're looking for
(24:16):
other air traffic. They're looking for sort of is the
system performing as we expect it to? Is there is
there weather that's rolling in? Is there something else like that?
Speaker 2 (24:24):
So just so I can see, just so I can
like visualize it, is there like at each place where
you have a nest, like at each Walmart, is there
like a room where there's a guy sitting or do
you guys have like an office.
Speaker 3 (24:36):
There's an office where the folks are sitting.
Speaker 2 (24:39):
I mean you said there's a pilot, Like do they
So what do they have like a little radar thing?
Speaker 3 (24:43):
Like pretty much like if you think about like sort
of what what an aircraftic controller station would look like,
or what a sort of train dispatcher station look like,
that's more or less what the pilot interface looks like.
So there is there's no live video feed from the aircraft.
So like the airplanes sort of get their assigned flight
when somebody places an order, because like we don't know
(25:04):
where the next order is going to come from. So
there's this whole route planning algorithm in the background that
when somebody orders something, it's like, okay, I need to
go fly this way, and here are the other flights
in the space. They'll see those tracks like so they'll see, okay,
I've got you know this many airplanes that are sort
of active right now, I've got this many airplanes that
(25:24):
are on the ground waiting getting assigned in order. And
they're just little icons moving around the screen. So it's
not like there's not some video feed. There's not like
and then they've got like a view of the individual nests.
So there's a little security camera looking down on it.
It's like, oh, okay, there's here all the airplanes, like
you know, so it really is it is that sort
(25:45):
of like supervisor interface. I think it's probably the easiest
way to think about it, rather than like there's no
like sort of controls and handles that people are manipulating
to fly the planes around.
Speaker 2 (25:57):
I mean, is there like a button they can push
if the plane needs to turn around or like whatever
bail in some way?
Speaker 3 (26:04):
Yeah, there's a button. So one of the things they
could do is they can pause the whole operation, so
they can say like Okay, you know we're outside of
our weather limits, or there's a airspace conflict that has
gone through. Stop receiving orders.
Speaker 2 (26:19):
That's the easy one, don't any more planes out.
Speaker 3 (26:22):
The other one is they have a they have a
button where they can tell the plane to land, and
that's basically the plane is following like a four D trajectory,
so it's got a like a three D path plus
time where it's like, Okay, I'm supposed to be in
this airspace at this time, and it'll it'll keep flying
(26:43):
on that path until it gets to a place that
sort of has been procedurally generated as like okay to land,
so like you know, open fields and like trees and
stuff like, not roads, not like sidewalks, and so the
plane will keep flying until it gets to one of
those places and it'll sort of slow down to a hover,
come down, and land and that those are really the
(27:05):
those are the two commands.
Speaker 2 (27:07):
How often does that one happen?
Speaker 3 (27:09):
Very land, Yeah, very infrequently. And I think it's more
I think it's more akin to like like if you
pulled your car over to the side of the road
with a check engine light it's it's not like it's
not like sort of anything is like fundamentally broken. It's
more like a precautionary step. And I think that's that's
(27:29):
one of the pieces that sort of underpins this whole
sort of operational cases, like all these layers of redundancy.
The plane can either the plane can go to the
delivery zone and say like, well, I don't think I
should deliver the package because there's an obstacle and they'll
come back. The plane can go and deliver the box
and say I delivered it, but I had to move
two meters over because there was an obstacle, like disable
(27:50):
this delivery zone. The plane can come back and say, hey,
I was doing that, and when I was on my
way back, I noticed that I was using slightly more
power than I usually use, like I'm gonna ground myself.
And sort of there's all these layers, and one of
those layers is the plane or the pilot can say, hey, like,
why don't you pull over the side of the road
(28:10):
so we can we could check on you.
Speaker 2 (28:14):
Uh is there a camera on the plane?
Speaker 3 (28:16):
Yeah, surely.
Speaker 2 (28:17):
Something many people ask you, like, is the drone looking
at me?
Speaker 3 (28:21):
So the drone is not looking at you. Again, I
think it's like there's no uh, there's no live feed,
so it's not like there's like an operator sort of
interfacing that video. But the aircraft has two sets of
stereo cameras on the back of it that are used
for uh so both visual odometry so like a backup
(28:42):
to GPS, and for like sort of broad scene understanding,
so like building a depth map of sort of what
you're flying over. But those are all like that's a
point cloud, that's not like images of stuff. You know.
One of the biggest challenges of like residential drone delivery
is like you're going to you're going from a very
(29:03):
structured environment to a very unstructured environment. So like like
the sky is like usually it's very clear, clear of obstacles.
Somebody's backyard like may not be clear obstacles. There may
be there may be a tree branch that has grown
sort of since the last time that sort of satellite
imagery was refreshed. And so the cameras on board the
(29:24):
aircraft are used mostly for the reason they're used for
navigation and for sort of obstacle detection. They run what's
what are what's called like a bunch of semantic modeling.
So like if you think about like a toddler putting
blocks into the little toy, they'll be like triangle goes
in triangle, circle goes in circle. The plane is like car, tree, house, road, huh.
Speaker 2 (29:49):
Like computer vision, like basic computer visions, computer vision stuff.
So that one is interesting in that I feel like
when you started that wasn't nearly as good.
Speaker 3 (30:01):
Right, Yeah, you can do We could do a lot
more now than then we could at the beginning of
the program. And like you know, we we can run
like live classification on like you know, cell phone like
avion X.
Speaker 2 (30:14):
Yeah, I mean, does does it also do that to
say like, oh, don't lower the box. There's the dog
running around the.
Speaker 3 (30:20):
Yard yep, yep, so'll it'll do that. And it'll do
basically what's what's called delivery zone nudging, so like a
spot is picked sort of before the plane gets there
where to lower the box. If the plane is coming down,
it says, well, where you told me to lower the box.
There's a car or the tree branch is bigger than
(30:40):
what you said it was. It can, within a bounded
set of decisions sort of move it a certain distance
to drop off the box and then it will tell
us like, hey, I had a move where the delivery
zone was.
Speaker 2 (30:53):
How often how often does a delivery fail? Like does
it get there and it's like, no, this isn't going
to work. I got to get out here pretty infrequently.
I mean, like one, it's in the hundreds or thousands. Okay,
So so big picture, if things go well for drone delivery,
what will the world look like in say five or
ten years.
Speaker 3 (31:12):
Yeah. I think at the core of it, if if
this experiment works right, like if if drne delivery takes off,
it will be the like primary means that people get
small goods delivered, So I think that it will be
it'll be everyday occurrence for the same folks that order
food every day today. Most of that will be coming
(31:33):
with airplanes. Because think it just just makes more sense.
Uh huh.
Speaker 2 (31:39):
What's the version of the future where it doesn't work? Like,
what are the reasons it might not work?
Speaker 3 (31:45):
I think some of it is if people can actually
get past the what ifs, like you know, do people
order frequently enough? Is it a good experience for the partners?
Like are you integrating well enough into their workflow? Where
Joon delivery is an asset, not a burden to their
sort of the day in the life of their associates.
(32:06):
Are all the companies in the space still around right?
Like this is like these are really long timelines in
a challenging funding environment.
Speaker 2 (32:14):
So like, you know, it's really expensive to get to
a point where the business makes sense to get to scale.
Speaker 3 (32:21):
And then I think I worry sometimes that that folks
will sty'll sort of chase the long tail of the
sort of use case. So you know, drones don't have
to deliver everything, right, and so I worry about sort
of a lack of focus of like okay, like you know,
(32:42):
here's a place where drones make a ton of sense
to deliver stuff. Let's be really successful there rather than like, oh,
they might be able to deliver you know, the case
of bowling balls, Like yeah, they might be able to,
but advan is a much better use case for that.
Speaker 1 (32:59):
We'll be back in a minute with the lightning round.
Speaker 2 (33:11):
So I just want to do a lightning round. I
appreciate your time. I want to talk about I want
to talk about your Instagram, Like it's it's stuff like
I'm looking and like here's like it's a lot of
Star Wars, like planes like X wing Fighter, and there's
the X wing Fighter and it's it's actually flying right
(33:33):
you click on it. And there's the like winged win
a Bago from space Balls. There's the flying toaster, which
is like a cut from the it's like the ninety
screen saver, right, and like so you take these things
and you actually make.
Speaker 3 (33:45):
Them fly by. So not surprisingly, I'm a bit of
a nerd, like I like star Wars and sci fi things.
I always have, and I reached a point where sort
of both the intersection of technologies and my skill at
designing airplanes were such that I could make the toys
(34:08):
that I had as a can work as like real things.
Speaker 2 (34:11):
Right, So you take the like plastic X wing fighter
that I had in my room, and like I made
it fly by like throwing it in my hand and
making the sounds right, and you actually make it fly.
Speaker 3 (34:22):
Yeah, no, it's it's It's also like how I sort
of continuously sharpened my skills as an engineer. So like
often they're like really hard airplane design problems of like
what are the aerodynamics of an X wing?
Speaker 2 (34:35):
So let me ask you on that note, let me
ask you, like is there a specific thing you learned
in building these models that you brought to your work
at Wing.
Speaker 3 (34:47):
I think I've brought a similar philosophy, which is like,
how do you how do you get to a complex
result without a complex solution? Where it's like, okay, how
do I how do I get that same joy? How
do I get to that same result with a simpler answer? So, uh,
that's I think that that's that's how I approach my
(35:07):
professional life as well.
Speaker 2 (35:09):
You got a white whale. Is there something you want
to build that flies that you haven't been able to do?
Speaker 3 (35:14):
I have always wanted to do a tie interceptor. This
is always my favorite star warship. But there's like not
much in the way of wing on that.
Speaker 2 (35:22):
Who flew that?
Speaker 3 (35:23):
Was that?
Speaker 2 (35:23):
The the good guys? Okay, right, so it's like what
the storm Troopers flew? Sort of a variation on that.
You built one of those that goes in the water
some some version of that.
Speaker 3 (35:36):
Yeah. I saw that.
Speaker 2 (35:37):
I was like, does that fly? And then I saw
that it was in the.
Speaker 3 (35:39):
Like every probably every four years, I get in the
submarines and I'll like build some submarine stuff and then
I'll put it on the shelf.
Speaker 2 (35:51):
So I feel like I'm I'm moving on from what
you built now. Just to be clear, the I feel
like the word drone is is not doing drones any favors, right,
I feel like it has a lot of negative notations.
There's like a militaristic connotation. There's the connotation of like
(36:12):
a worker who is unfeeling and doesn't do anything. Is
there a word that you think would be better?
Speaker 3 (36:19):
I mean at this point, I think that that that
ship is long since sailed in his left board. I mean,
there was like a there's been a South Park episode
about it, Like I don't think that like like uh so,
I don't know. I think it's it's a word. I
don't think you should ever be afraid of of of
the word. And at the core of it, like I
(36:40):
don't know, I think of them all as airplanes, Like
they fly around with other airplanes, like they need to
do airplane stuff. And I think that like part of
the reason that the skepticism that that you brought up
earlier exists is the sort of the gap between the
sort of promise and the reality. And so you know,
the the sooner that we can get these services sort
(37:03):
of real and live and interacting with more people. I
think that people will just look at these as tools
that are flying around providing a service. And my I've
always I tell the team frequently, I like, I want
people to look at the plane and think of it
like a shopping cart, and it's just like it's a
thing that's very useful. People use them all the time,
(37:25):
but you don't really give it a second thought, and
it's just that it's an asset that you use.
Speaker 2 (37:32):
Adam Woodworth is the CEO of Wing. Today's show was
produced by Gabriel Hunter Chang. It was edited by Lydia
Jean Kott and engineered.
Speaker 3 (37:41):
By Sarah Bruguer.
Speaker 2 (37:42):
Please email us we are at problem at pushkin dot fm.
Let us know would interview on the show, how to
make the show better, et cetera. I'm Jacob Boldstein and
we'll be back next week with another episode of What's
Your Problem
Pushkin.
Speaker 2 (00:20):
If you live in the Dallas Fort Worth area, there
is a good chance that right now you can open
up an app on your phone and order a cup
of coffee or a bottle of ibuprofen from a nearby Walmart,
and then about fifteen minutes later, a drone will appear
hovering in the sky above your yard. The drone will
lower down a rope that holds a box that has
(00:43):
your ibuprofen in it. Then the drone will leave the
box on the ground, pull up the rope and fly away.
People have been talking about drone delivery for a long
time now. There have been pilot programs and test sites,
but this thing in Dallas is big. Millions of people
live in the delivery zone. It's the first place in
(01:03):
the US where for lots and lots of people, drone
delivery seems set to become a normal, everyday thing. I'm
Jacob Goldstein and this is What's Your Problem, the show
where I talk to people who are trying to make
technological progress. My guest today is Adam Woodworth. Adam is
(01:26):
the CEO of Wing. Wing is owned by Alphabet and
it's one of the two drone companies doing deliveries for
Walmart in Dallas. Wing has also been making deliveries in
Australia for several years now. Adam's problem is this, how
do you turn a delivery drone into something as boring
and ubiquitous as a shopping cart? To start, I asked
(01:48):
Adam to make the case for his project, like, why
should I be happy to learn that delivery drones may
soon be coming to a sky near me delivering coffee
and burritos to my neighbors.
Speaker 3 (02:01):
I've caught a lot of flack over the years for
the burrito copter, and like, oh, you're just using drones
to fire around and deliver cups coffee and stuff. But
at the core of it, like people eat food every day,
Like if you look at it from a holistic transportation perspective,
like people are not going to stop ordering things, Like
(02:22):
that's not like it's in fact the opposite. People are
ordering more stuff every day. And it just does not
make sense that we are moving around like burritos and
cups of coffee in cars, Like it just does not
make sense.
Speaker 2 (02:38):
To what extent are people ordering coffee delivered by somebody
in a car? Is that a thing?
Speaker 3 (02:42):
And I just don't know it. It is a thing.
It is a thing like you can go sort of
look up. There's been a myriad of different studies about
sort of what actual sort of like dipler use cases are,
and it's a lot of small items. I can I
can talk about a bunch of aviation math, but I
think the simplest way to look at it is we
are designing systems like to be able to fly in
(03:04):
the national airspace. This is the same for everybody who's
sort of in here to be building systems that are
like sort of equivalent level of safety to like airliners.
So if you take that and you abstract it down
to things that people can sort of internalize in terms
(03:24):
of sort of risk, it is safer to fly the
package via drone than it is to walk the package
in your hand down the sidewalk.
Speaker 2 (03:36):
That's good, that's compelling, that's fun. Non intuitive, it's very
non intuitive. I thought you were gonna say safer than
to drive, and I'd be like, sure.
Speaker 3 (03:43):
Everything everything, but like safer than to hand carry it.
Speaker 2 (03:47):
The risk of hand caring is getting hit by a car.
Speaker 3 (03:50):
Yes, that's it's usually serve pedestrian road risk. And so
you have this form of transportation which is possible and
real and financially viable. That is many orders of magnitude
safer than the incumbent. If you look at on every
metric from safety to emissions, sort of impact on infrastructure,
(04:11):
It's like, it is silly that we are moving around
these boxes and the way we do today. This is
a fundamentally better way to do it.
Speaker 2 (04:18):
So if I want to order something, I live near
Walmart in Dallas and I want to order something from you, Like,
how does that part work?
Speaker 3 (04:25):
So basically like we built a pretty traditional looking marketplace
where the order flow is very traditional until you get
to the part at the end, which is the drone part.
Speaker 2 (04:37):
So I use your app. I don't use like the
Walmart app.
Speaker 3 (04:39):
So it depends on which region you're in. So in
Australia we're integrated into the door Dash app, So use
the front end of door Dash. In the US right
now we're integrate. We're using our own app. Okay, So
you say what you want ordered, and then before you
even get to that step, the back end will say
is your address deliverable? Right? Like it is the place
(05:01):
that you are, a place that the plane can get
to just to be.
Speaker 2 (05:04):
Clear that is you checking it out. You're using basically
satellite data to say does this address work? Or that's
you asking music. No, that's you just checking.
Speaker 3 (05:12):
So it's just basic. And some of it is even
like like if there's a big hill that the plane
has to spend more energy climbing over, like you might
be within the radius, but like your particular house like
may be undeliverable, or you might be slightly outside of
the radius because there's no there's no power lines or
hills to get there, so the plane go slightly farther.
(05:34):
So everybody thinks about a delivery area as a circle,
but it's really like a weird amiba.
Speaker 2 (05:41):
With lots of variables affecting where you can go.
Speaker 3 (05:43):
Yeah, and so there's a there's a sort of there's
an address checker that'll run that like just GPS coordinate
through a terrain database to be like can the plane
get there? And then once you get to check out,
you'll be presented with either one delivery zone of like okay,
this is the place on your property where the box
can be dropped off, or a choice like if there's
(06:04):
multiple spots, it'll be like okay, which one of these three,
do you want it in the left side of your
yards sell the ardor in the backyard the personal pick that.
Then the order will process, the plane will be dispatched,
and then it'll go to like a sort of countdown
of like, Okay, the plane is on its way and
it'll be here in three minutes and forty two seconds.
Speaker 2 (06:25):
So uh, and and what's the cost?
Speaker 3 (06:29):
It depends it's uh, it's sort of uh. That's that's
that's more of a partner question.
Speaker 2 (06:35):
What's it what's it cost you?
Speaker 3 (06:37):
That's a that's an unanswerable question.
Speaker 2 (06:40):
A lot right, now, A lot right?
Speaker 3 (06:43):
No, I mean I think, uh, I think less than
less than what folks would assume.
Speaker 2 (06:49):
I mean, the marginal cost is low, right, but it's
not that many deliveries against a large fixed costs.
Speaker 3 (06:54):
That's fair. But the marginal metrics are all sort of
you know, favorable or in the direction of favorable. Yeah. So,
like you know, people are well accustomed to getting stuff
delivered today, and so like those costs are well understood,
and the sort of costs per delivery that you know,
(07:14):
either is either the partner is charging they're sort of
all in line with that, depending on sort of which
region you're in. But that's that's ultimately up to the partner.
Speaker 2 (07:23):
A few bucks, I mean, is is that what that means?
Reason which you know it can't be more than that
if people are out in coffee, right though, it is
amazing how much people spend on, you know, on the
marginal thing. So so that's how it works from the
consumer's point of view, How does it work from you know,
from the drones point of view. From from the company's point.
Speaker 3 (07:42):
Of view, what happens every every morning the service turns on,
Folks can can order the stuff. Planes will get automatically
tasked to go take off, pick up the box and go.
Speaker 2 (07:54):
How many deliveries what they do in a day?
Speaker 3 (07:58):
Like our high water mark right now is around for
one store was about like one hundred and thirty deliveries, okay,
So it's like these are like, these are real numbers.
This isn't like like, oh okay, like three people ordered
stuff today, And I think that that's the that's I
think sort of the phase that the industry is moving
into now is it's it's these aren't like pilot programs
(08:21):
and trials. These are like, okay, this store opened up
offering this service and all the people in the area
expect to get that service.
Speaker 2 (08:28):
How how big and how heavy can something be?
Speaker 3 (08:31):
So you know, we carry about a little bit over
a kilo of of of good, so you know, two
point two pounds of stuff, and in.
Speaker 2 (08:39):
Terms of volume, like.
Speaker 3 (08:41):
Like I think about like a shoe box, shoe box
ish shoe box, okay, And a lot of it ends
up being you know, prepared food, drinks like you know,
soft drinks, you know, grocery items. And I think the
one of the interesting observations I've always had on this is, uh,
back back to the sort of original like did we
(09:03):
pick the right use case piece. We did a bunch
of surveys sort of early on in the program of
you know, what would you want delivered by drone if
you if you had the option, and you know, there
were dozens of different options sort of from food all
the way to medicine, clothing, all these other things, and
overwhelming the responses were always like, you know, drones should
(09:26):
only be used for medical delivery, like they should only
be used for medicine first aid bits like that. And
then we went and offered the service in Australia. We
had we had all those goods in there. We had
like a lot of you know, like type of you know,
over the counters, pharmacy items, sort of food, clothing, a
(09:48):
bunch of different local restaurants, and like eighty percent of
people ordered coffee. Like that was the overwhelming sort of
thing because it's like that's something that people got every day,
and it made.
Speaker 2 (09:59):
Sense stated preference versus revealed preference. I mean there is
there is also the externality thing, right, Like it's one
thing to say, if a drone's going to be flying
over my house all the time, what do I want
it to be delivering to everybody? Versus if it's coming
to my house, what am I going to order today?
Speaker 3 (10:16):
Right? I mean that is another way to frame that's fair.
Speaker 2 (10:19):
So like what what are when you look back after
ten years of working at it and like now you're
actually doing it, Like, what are a few key things
you figured out?
Speaker 3 (10:30):
I think one of the most defining moments is or
maybe one of the most defining constraints. Most sort of
disruptive or merging technologies sort of start off in a
permissive environment. So you know, one of the one of
the things that I you know, one of the questions
I've gotten asked many times sort of over the years,
(10:52):
is well, why didn't you take the approach that like
ride share took of. Okay, you know, offer the service
in a big metro area, build up a big customer base,
and then as the rules change, you have, like you
have customers to say, Okay, this is a valuable thing.
Speaker 2 (11:08):
Why didn't you do it?
Speaker 3 (11:10):
Why not just go? Why not why not just go?
Do it? And I think at the core of it,
it's because aviation does not operate that way right, Like
aviation is is is built around a preempted set of
rules where the federal government controls the airspace, so it's
not up to sort of anyone given municipality. And from
that place, you know, how do you build consumer demand,
(11:35):
How do you show that this thing works? How do
you even go and do a trial where the existing
regulatory frameworks are just say no, you can't fly at all.
Speaker 2 (11:44):
Yes to Australia.
Speaker 3 (11:47):
Yeah, So that's that's what That's what drove us to
to start in Australia because they were one of the
first markets where the National Aviation Regulator there they sort
of quantify the risk of the operation. They look at
the specifics the operation and they sort of approve on
a case by case basis of all right, like this
meets the same level of safety as an airliner, even
(12:09):
though it's a very different use case.
Speaker 2 (12:12):
So just tell me like one thing that you know
you had to figure out. Maybe you tried something and
it didn't work, and you tried something else and it worked.
You know what I mean? Is there just one kind
of a microcosm example.
Speaker 3 (12:24):
If I could do two a lot? If that's yeah.
So the first one is sort of how do you
achieve aerospace reliability with like consumer electronics costs? So you know,
we started from a place of how do you build
these planes in a way where the unit economics will work?
(12:45):
Not how do you build these planes and then try
to figure out how to make the unit economics work?
Speaker 2 (12:49):
Right?
Speaker 3 (12:50):
So what is I mean?
Speaker 2 (12:51):
Obviously the cost of one is going to be a
function partly or largely of scale. But since you brought
it up, like, what has one of them got a
cost for it to work for you?
Speaker 3 (13:00):
I mean it has to look like cell phones basically.
And I think that the interesting one that the sort
of specific point here is is if you looked at
just the reliability case Okay, so how many propulsion system
elements do you need in order to be able to
make the airplane sort of close all the reliability numbers.
(13:21):
That requires eight motors in our use case, so you know,
you need eight hover motors so that you can have
any one of them fail at any given time and
the airplane can still sort of have controlled flight. In
our cost trades, it ended up being that if you
move to twelve motors instead of eight motors, so the
(13:43):
reliability numbers close at eight. But if you went to twelve,
you got to use a much smaller motor that didn't
need a carbon fiber propeller. You could use a plastic propeller.
And there was a huge change in the cost of
the total propulsion system by increasing the number of motors,
which is not a thing that you would on the
surface look at and be like, okay, this is intuitively obvious.
Speaker 2 (14:08):
So you're using twelve cheaper motors, so more smaller. It's
more smaller motors rather than fewer larger motors. And I
mean presumably you could keep going that way. You could
have twenty tiny motors. I mean, you find the optimum.
Speaker 3 (14:25):
What settled on twelve was really driven by noise. So
like as you get the burders smaller, they have the
you know, the frequencies go higher, they spin faster, that
discon goes up all these different effects. And so moving
to twelve let you get the sort of cost benefits
by going small motors. But you didn't sort of start
(14:47):
climbing up that that noise curve by having like you know,
like one hundred little half inch motors hu and even
that there isn't an off the shelf propeller solution that
sounds pleasant.
Speaker 2 (15:02):
Yeah, So that the sound, the sound is a big one, right,
Like there was a good Wall Street Journal story I thought,
uh from Australia where you were doing a lot of
the early work, and the headline was delivery drones, cheer shoppers,
annoy neighbors, scare dogs. Tell me about that piece of it.
Tell me about the noise.
Speaker 3 (15:22):
So that's a that's a it's a thing you have
to be conscious of. So, you know, we did a
bunch of work early on in the development of this
aircraft around sort of how do we how do we
shape the tone of the noises being generated. So it's
sort of less of like one big peak and sort
of more broad spectrum, how do we reduce the noise
of it? Like, if you look closely at our propellers,
(15:44):
they're they're like asymmetrics. So there's one long set of
blades and one short set of blades, and that creates
like two tip forto sees that destructively interfere and like
spread out noise. So all these things. So we did
all this work on on the hover propulsion system, and
the vast majority of of sort of noise feedback from
the communities and mind you, like, if you compare this
(16:07):
to the noise feedback you get about an actual airport
or you know, construction project is still like far less
than that. But it was all overflight, so it was
like people weren't complaining about the airplanes hovering. They were
complaining about them flying over overhead. And if you you
(16:27):
look at the fact that the person that ordered the goods,
like the plane is coming, then they asked for it, right,
Like they're like they're like, I want the service. Their
neighbors are probably also ordering things. But the thousand homes
that you flew over to get to that spot, like
may not be participating in the service.
Speaker 2 (16:45):
It's a negative externality, right, you're imposing a cost on
someone who's not a party to the transaction.
Speaker 3 (16:51):
So so that was that was a very interesting learning
sort of early on in our that we only would
have learned by actually operating the service.
Speaker 2 (16:59):
Huh, it's that it's not it's not the next door
neighbor who hates it. It's all the people between the
store and the customer.
Speaker 3 (17:06):
And in that sense, it was like, okay, well, well,
like we thought we needed to put all this development
work in the hover propulsion system, but it's actually the
crew system that they need to work.
Speaker 2 (17:16):
People never get mad about what you think they're going
to get mad about, right again, I.
Speaker 3 (17:20):
Think that that's the at the core of it, Like
the dron delivery space is like so full of has
been so full of what ifs for so long, and
in the absence of regulatory frameworks that let you go
and operate like you can't move out these things that
it it oftentimes like there's a lot of misconceptions about
(17:41):
either noise or sort of what the use cases will
be or what people want to order, and then when
you go and offer the service, sort of the cure
for the what ifs is the actuality of an operating.
Speaker 2 (17:52):
So okay, so you realize that the noise complaints aren't
the noise complaints you're expecting. You had mitigated the one,
the hovering, but not the other, the flying over. So
now you have this new challenge, right that people getting
mad if a drones are flying over their house all
day long? What do you do about that?
Speaker 3 (18:08):
So we we basically did the same thing we did
on the other propulsion side, So we used the same
technology as to redesign the props. We sort of set
the design target so that like basically the aircraft that
the cruise altitude would be inaudible to a to an
observer on the ground. But this is like everything in
aerospace is a trade and so you know, for us,
(18:32):
it was like we want to we want to address
sort of any of the noise concerns. Uh, And so
we're willing to go and do that development time and
exercise and costs, redo the propulsions, and then redo the
plane to to be able to use it. So that
that's uh, that was that was one interesting one and
(18:54):
the other one is just like you know, how you
get the box to the people. So early on in
the program, uh, we had basically you know, you have
the little tope that hangs below the plane, and then
it was connected to playing with a tether, and that
tether went around a school and just sort of like
you'd open up a latch and it would descend and
(19:17):
you'd leave behind the string. So it was like the
way of letting go the box was just well, there's
only so much string on the on the on the aircraft.
Speaker 2 (19:26):
So so the customer would get a box with a
long string.
Speaker 3 (19:30):
Not a great user experience, but very simple, like super simple.
And uh, we spent a huge amount of time around
how do you drop off the box without a bunch
of mechanical complexity? So how do you how do you
lower down the box, make sure that it lets go
(19:50):
of the box one hundred percent of the time. Make
sure that that little hook at the bottom doesn't get
snagged on anything else as it's getting pulled back up. Uh.
Making sure that that whole system could operate reliably, you know,
thousands of times with dirt and leaves and stuff. Uh.
And so like this is little, that little yellow bit
(20:11):
of plastic. So you're just.
Speaker 2 (20:13):
Talking about the hook at the end of the line
that holds the box. Like that is like a nightmare
to get.
Speaker 3 (20:19):
Just right, nightmare to get just right. And it is
probably one of the most iterated components on the whole aircraft.
Speaker 2 (20:27):
Still to come on the show. Why despite all that
work on the hook and on everything else, drone delivery
may never really take off. Also, what the world will
look like if drone delivery does work? So what like
(20:53):
what's the frontier for you? Like, what are you trying
to figure out that you haven't figured out yet?
Speaker 3 (20:59):
Yeah, I think that for us right now. A couple
of years ago in Australia when we were sort of
ramping up volumes there, we demonstrated what scale looked like
for a sort of confined geographic area. So, you know,
one of the common I think misconceptions about drones is,
(21:19):
you know, folks like, well, when drone delivery is real,
I'll look up at the sky and all I see
is drones. And I was like, well, no, when you
when you actually do it, when you actually offer sort
of the scaled out service to like a you know,
one hundred square kilometer piece of the Earth, even when
there's even then when there's a thousand deliveries happening in
one day, and that that little space you can look
(21:41):
up and not see any planes for a while. And
so for the last couple of years that's what we
were focused on was Okay, how do we demonstrate the
unit scale of Okay, this is what drune delivery looks
like at scale for one one small space. Now I
think with the DFW expansion, we have the opportunity to
(22:02):
do that at a whole sort of metro wide scale.
So I think that the ability to offer the service
across a big metro to like, you know, where you're
gonna cover millions of people. Let's you answer, let's you
sort of validate that those assumptions that have been proven
out at smaller operational scales hold as you go up
(22:23):
and so looking at things like how frequently do people order,
like is it is it a novelty that people order
once and then they're like, okay, well I got my
cool TikTok video of a drown delivering me something or
do they find value in it? You know, do you
start to get to the place where you know, for us, like,
you know, is one operator able to you know, do
(22:47):
we have the permissions and is it working to have
one operator do multiple locations and you know, many aircraft
in the sky at once looking less and less like
sort of a one to one relationship between sort of
the person the operation to like more of a sort
of dispatcher view of the world of like, okay, this
person is managing a fleet of airplanes rather than energing
(23:09):
sort of anyone flight.
Speaker 2 (23:11):
So in the in the current service as it exists, Now,
what what is the role of human beings?
Speaker 3 (23:19):
It's it's really is like, uh, you know, a person
has to put the stuff in the box. A person
has to bring the box to where the plane is
going to be, and you know, the plane hovers and
lowers down the little hook and they hook the box
on it and it takes it up. And then you know,
somebody in the morning needs to take the planes and
put them out on the pads, and at night take
(23:41):
them and put them back in the shipping container. And
then our pilots are sort of monitoring the operation.
Speaker 2 (23:46):
I mean, just to be clear, they're not flying the planes, right.
The planes are flying themselves.
Speaker 3 (23:51):
Planes are the planes are flying themselves, but they're responsible
for the operation, right. So like if if you look
at the sort of like some of the core tenets
of aviation, like the pilot is responsible for the airliner
flight or for the charter flight. Like they're responsible for
the operation. And so what does that mean when things
get very highly automated, it's okay, well they're looking for
(24:16):
other air traffic. They're looking for sort of is the
system performing as we expect it to? Is there is
there weather that's rolling in? Is there something else like that?
Speaker 2 (24:24):
So just so I can see, just so I can
like visualize it, is there like at each place where
you have a nest, like at each Walmart, is there
like a room where there's a guy sitting or do
you guys have like an office.
Speaker 3 (24:36):
There's an office where the folks are sitting.
Speaker 2 (24:39):
I mean you said there's a pilot, Like do they
So what do they have like a little radar thing?
Speaker 3 (24:43):
Like pretty much like if you think about like sort
of what what an aircraftic controller station would look like,
or what a sort of train dispatcher station look like,
that's more or less what the pilot interface looks like.
So there is there's no live video feed from the aircraft.
So like the airplanes sort of get their assigned flight
when somebody places an order, because like we don't know
(25:04):
where the next order is going to come from. So
there's this whole route planning algorithm in the background that
when somebody orders something, it's like, okay, I need to
go fly this way, and here are the other flights
in the space. They'll see those tracks like so they'll see, okay,
I've got you know this many airplanes that are sort
of active right now, I've got this many airplanes that
(25:24):
are on the ground waiting getting assigned in order. And
they're just little icons moving around the screen. So it's
not like there's not some video feed. There's not like
and then they've got like a view of the individual nests.
So there's a little security camera looking down on it.
It's like, oh, okay, there's here all the airplanes, like
you know, so it really is it is that sort
(25:45):
of like supervisor interface. I think it's probably the easiest
way to think about it, rather than like there's no
like sort of controls and handles that people are manipulating
to fly the planes around.
Speaker 2 (25:57):
I mean, is there like a button they can push
if the plane needs to turn around or like whatever
bail in some way?
Speaker 3 (26:04):
Yeah, there's a button. So one of the things they
could do is they can pause the whole operation, so
they can say like Okay, you know we're outside of
our weather limits, or there's a airspace conflict that has
gone through. Stop receiving orders.
Speaker 2 (26:19):
That's the easy one, don't any more planes out.
Speaker 3 (26:22):
The other one is they have a they have a
button where they can tell the plane to land, and
that's basically the plane is following like a four D trajectory,
so it's got a like a three D path plus
time where it's like, Okay, I'm supposed to be in
this airspace at this time, and it'll it'll keep flying
(26:43):
on that path until it gets to a place that
sort of has been procedurally generated as like okay to land,
so like you know, open fields and like trees and
stuff like, not roads, not like sidewalks, and so the
plane will keep flying until it gets to one of
those places and it'll sort of slow down to a hover,
come down, and land and that those are really the
(27:05):
those are the two commands.
Speaker 2 (27:07):
How often does that one happen?
Speaker 3 (27:09):
Very land, Yeah, very infrequently. And I think it's more
I think it's more akin to like like if you
pulled your car over to the side of the road
with a check engine light it's it's not like it's
not like sort of anything is like fundamentally broken. It's
more like a precautionary step. And I think that's that's
(27:29):
one of the pieces that sort of underpins this whole
sort of operational cases, like all these layers of redundancy.
The plane can either the plane can go to the
delivery zone and say like, well, I don't think I
should deliver the package because there's an obstacle and they'll
come back. The plane can go and deliver the box
and say I delivered it, but I had to move
two meters over because there was an obstacle, like disable
(27:50):
this delivery zone. The plane can come back and say, hey,
I was doing that, and when I was on my
way back, I noticed that I was using slightly more
power than I usually use, like I'm gonna ground myself.
And sort of there's all these layers, and one of
those layers is the plane or the pilot can say, hey, like,
why don't you pull over the side of the road
(28:10):
so we can we could check on you.
Speaker 2 (28:14):
Uh is there a camera on the plane?
Speaker 3 (28:16):
Yeah, surely.
Speaker 2 (28:17):
Something many people ask you, like, is the drone looking
at me?
Speaker 3 (28:21):
So the drone is not looking at you. Again, I
think it's like there's no uh, there's no live feed,
so it's not like there's like an operator sort of
interfacing that video. But the aircraft has two sets of
stereo cameras on the back of it that are used
for uh so both visual odometry so like a backup
(28:42):
to GPS, and for like sort of broad scene understanding,
so like building a depth map of sort of what
you're flying over. But those are all like that's a
point cloud, that's not like images of stuff. You know.
One of the biggest challenges of like residential drone delivery
is like you're going to you're going from a very
(29:03):
structured environment to a very unstructured environment. So like like
the sky is like usually it's very clear, clear of obstacles.
Somebody's backyard like may not be clear obstacles. There may
be there may be a tree branch that has grown
sort of since the last time that sort of satellite
imagery was refreshed. And so the cameras on board the
(29:24):
aircraft are used mostly for the reason they're used for
navigation and for sort of obstacle detection. They run what's
what are what's called like a bunch of semantic modeling.
So like if you think about like a toddler putting
blocks into the little toy, they'll be like triangle goes
in triangle, circle goes in circle. The plane is like car, tree, house, road, huh.
Speaker 2 (29:49):
Like computer vision, like basic computer visions, computer vision stuff.
So that one is interesting in that I feel like
when you started that wasn't nearly as good.
Speaker 3 (30:01):
Right, Yeah, you can do We could do a lot
more now than then we could at the beginning of
the program. And like you know, we we can run
like live classification on like you know, cell phone like
avion X.
Speaker 2 (30:14):
Yeah, I mean, does does it also do that to
say like, oh, don't lower the box. There's the dog
running around the.
Speaker 3 (30:20):
Yard yep, yep, so'll it'll do that. And it'll do
basically what's what's called delivery zone nudging, so like a
spot is picked sort of before the plane gets there
where to lower the box. If the plane is coming down,
it says, well, where you told me to lower the box.
There's a car or the tree branch is bigger than
(30:40):
what you said it was. It can, within a bounded
set of decisions sort of move it a certain distance
to drop off the box and then it will tell
us like, hey, I had a move where the delivery
zone was.
Speaker 2 (30:53):
How often how often does a delivery fail? Like does
it get there and it's like, no, this isn't going
to work. I got to get out here pretty infrequently.
I mean, like one, it's in the hundreds or thousands. Okay,
So so big picture, if things go well for drone delivery,
what will the world look like in say five or
ten years.
Speaker 3 (31:12):
Yeah. I think at the core of it, if if
this experiment works right, like if if drne delivery takes off,
it will be the like primary means that people get
small goods delivered, So I think that it will be
it'll be everyday occurrence for the same folks that order
food every day today. Most of that will be coming
(31:33):
with airplanes. Because think it just just makes more sense.
Uh huh.
Speaker 2 (31:39):
What's the version of the future where it doesn't work? Like,
what are the reasons it might not work?
Speaker 3 (31:45):
I think some of it is if people can actually
get past the what ifs, like you know, do people
order frequently enough? Is it a good experience for the partners?
Like are you integrating well enough into their workflow? Where
Joon delivery is an asset, not a burden to their
sort of the day in the life of their associates.
(32:06):
Are all the companies in the space still around right?
Like this is like these are really long timelines in
a challenging funding environment.
Speaker 2 (32:14):
So like, you know, it's really expensive to get to
a point where the business makes sense to get to scale.
Speaker 3 (32:21):
And then I think I worry sometimes that that folks
will sty'll sort of chase the long tail of the
sort of use case. So you know, drones don't have
to deliver everything, right, and so I worry about sort
of a lack of focus of like okay, like you know,
(32:42):
here's a place where drones make a ton of sense
to deliver stuff. Let's be really successful there rather than like, oh,
they might be able to deliver you know, the case
of bowling balls, Like yeah, they might be able to,
but advan is a much better use case for that.
Speaker 1 (32:59):
We'll be back in a minute with the lightning round.
Speaker 2 (33:11):
So I just want to do a lightning round. I
appreciate your time. I want to talk about I want
to talk about your Instagram, Like it's it's stuff like
I'm looking and like here's like it's a lot of
Star Wars, like planes like X wing Fighter, and there's
the X wing Fighter and it's it's actually flying right
(33:33):
you click on it. And there's the like winged win
a Bago from space Balls. There's the flying toaster, which
is like a cut from the it's like the ninety
screen saver, right, and like so you take these things
and you actually make.
Speaker 3 (33:45):
Them fly by. So not surprisingly, I'm a bit of
a nerd, like I like star Wars and sci fi things.
I always have, and I reached a point where sort
of both the intersection of technologies and my skill at
designing airplanes were such that I could make the toys
(34:08):
that I had as a can work as like real things.
Speaker 2 (34:11):
Right, So you take the like plastic X wing fighter
that I had in my room, and like I made
it fly by like throwing it in my hand and
making the sounds right, and you actually make it fly.
Speaker 3 (34:22):
Yeah, no, it's it's It's also like how I sort
of continuously sharpened my skills as an engineer. So like
often they're like really hard airplane design problems of like
what are the aerodynamics of an X wing?
Speaker 2 (34:35):
So let me ask you on that note, let me
ask you, like is there a specific thing you learned
in building these models that you brought to your work
at Wing.
Speaker 3 (34:47):
I think I've brought a similar philosophy, which is like,
how do you how do you get to a complex
result without a complex solution? Where it's like, okay, how
do I how do I get that same joy? How
do I get to that same result with a simpler answer? So, uh,
that's I think that that's that's how I approach my
(35:07):
professional life as well.
Speaker 2 (35:09):
You got a white whale. Is there something you want
to build that flies that you haven't been able to do?
Speaker 3 (35:14):
I have always wanted to do a tie interceptor. This
is always my favorite star warship. But there's like not
much in the way of wing on that.
Speaker 2 (35:22):
Who flew that?
Speaker 3 (35:23):
Was that?
Speaker 2 (35:23):
The the good guys? Okay, right, so it's like what
the storm Troopers flew? Sort of a variation on that.
You built one of those that goes in the water
some some version of that.
Speaker 3 (35:36):
Yeah. I saw that.
Speaker 2 (35:37):
I was like, does that fly? And then I saw
that it was in the.
Speaker 3 (35:39):
Like every probably every four years, I get in the
submarines and I'll like build some submarine stuff and then
I'll put it on the shelf.
Speaker 2 (35:51):
So I feel like I'm I'm moving on from what
you built now. Just to be clear, the I feel
like the word drone is is not doing drones any favors, right,
I feel like it has a lot of negative notations.
There's like a militaristic connotation. There's the connotation of like
(36:12):
a worker who is unfeeling and doesn't do anything. Is
there a word that you think would be better?
Speaker 3 (36:19):
I mean at this point, I think that that that
ship is long since sailed in his left board. I mean,
there was like a there's been a South Park episode
about it, Like I don't think that like like uh so,
I don't know. I think it's it's a word. I
don't think you should ever be afraid of of of
the word. And at the core of it, like I
(36:40):
don't know, I think of them all as airplanes, Like
they fly around with other airplanes, like they need to
do airplane stuff. And I think that like part of
the reason that the skepticism that that you brought up
earlier exists is the sort of the gap between the
sort of promise and the reality. And so you know,
the the sooner that we can get these services sort
(37:03):
of real and live and interacting with more people. I
think that people will just look at these as tools
that are flying around providing a service. And my I've
always I tell the team frequently, I like, I want
people to look at the plane and think of it
like a shopping cart, and it's just like it's a
thing that's very useful. People use them all the time,
(37:25):
but you don't really give it a second thought, and
it's just that it's an asset that you use.
Speaker 2 (37:32):
Adam Woodworth is the CEO of Wing. Today's show was
produced by Gabriel Hunter Chang. It was edited by Lydia
Jean Kott and engineered.
Speaker 3 (37:41):
By Sarah Bruguer.
Speaker 2 (37:42):
Please email us we are at problem at pushkin dot fm.
Let us know would interview on the show, how to
make the show better, et cetera. I'm Jacob Boldstein and
we'll be back next week with another episode of What's
Your Problem
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Nikki Glaser Podcast
Every week comedian and infamous roaster Nikki Glaser provides a fun, fast-paced, and brutally honest look into current pop-culture and her own personal life.