Day 217 in the Terra Hab.
It's getting pretty bad out there.
Winds are gusting 170 mph.
Global temperatures have risen two degrees in the last decade.
Tsunami swells are wiping out cities.
I'm just here trying to survive...
Long enough to...
-beat my high score... -[thunder crashes]
...and it looks like... I'm driving at the goal...
What, did we lose the genny?
You ever get the feeling we're whiling away the hours
as disaster looms?
Fiddling while Rome burns?
We're good on snow, Trev.
Some say we've got the tech
to right the planet within decades,
but what if we can't cool our climate,
or feed a population
that's projected to be 10 billion
by the year 2050?
-These topics makes me anxious. -[squirting sanitizer]
Thankfully, there's folks out there using A.I.
to tackle some of our most pressing problems,
like innovating agriculture
and redefining how we think about food.
And, oh, yeah, if things really do go south,
building a new home...
on another planet.
How can A.I.
help humanity adapt to a new way of life?
[man] Anybody on the arena floor
who is not a part of the competition,
please come inside as they prepare to do the testing.
[man 2] When I plugged it in the first time,
I plugged it into this.
-That's when the breaker blew. -Yeah.
Then I changed over to this cord.
[man 3] That's good.
It's feeling good, it's feeling good.
The 3D-printed habitat challenge
began as a call from NASA
asking anyone in the world from any background
to contribute architectural concepts
that you could make on Mars.
[Downey] Mars is the next great frontier.
NASA has been sending probes since the 1970s,
and, more recently, they even sent a space SUV
to explore the red planet.
[man over radio] Touchdown confirmed. We're safe on Mars.
[Downey] For decades, they've been laying the groundwork
for their real mission...
By 2033, they plan to send humans to Mars.
[Lynn Rothschild] To go to Mars,
it is about six months to get there,
and then you have to wait
about a year and a half on the surface
for the planets to realign for a very short trip back.
The Apollo 11 astronauts were in a tiny capsule.
[man over radio] Stand by for T-1.
[Rothschild] It had the volume on the lunar module
the equivalent of a C-block cell at Alcatraz
if it was only five foot high,
but you can't expect someone to live in a habitat like that
over any period of time.
[Downey] Um, yeah,
which is why they're gonna need a proper place to live
on the red planet.
The rub is it'll have to be built by A.I.
before they arrive.
[man] We have a few minutes,
if you could find a couple more cords
so we can get power for our electronics.
-[man 2] Okay, good. -[man 3] Good? Stop.
[Downey] NASA has been holding
an open bake-off for architects and engineers
to design a habitat
that can be built autonomously on Mars.
[Montes] Ready, Chris?
[Downey] Now, the competition has reached the final round,
with the winners getting half a million bucks
and a chance to join Project Red Planet.
[Montes] I don't know how many space architects
there are in the world.
It's not a common job.
There's not a whole lot of work yet, but...
my wildest dream is to get us into space faster,
but also revolutionize how things are built on Earth.
[man] So for Mars, we know we have to do
the pressure vessel shape, right?
And, Jeff, you had investigated,
-um, dome... -[Montes] Right.
[David Malott] I'm an architect.
My specialty is doing skyscrapers.
I want my company to be the first builders in space.
When you're building in space,
like, every little grain of material
is gonna have a value to it,
so we want something
that's gonna be as strong as possible
with as less material as possible.
Like, based on what you guys just printed,
what do you think is actually the easier build?
[Montes] We can kinda do whatever,
as long as we have enough time to lay down the plastic.
[Downey] In the past,
you'd have to use experience and empirical evidence
to figure out what to build,
which is why A.I. is so transformative
in the field of space exploration.
A.I. allows you to study millions of options,
comparing iterations against each other
and find the optimal one.
[Downey] This process
is sometimes known as "generative design,"
a fancy way of saying
machines learn to do what architects do.
Here, A.I. algorithms compare and predict
how different designs and forms function
in a virtual Martian environment.
[Montes] That function could be structural strength,
it could be thermal insulation,
or you could just be trying
to reduce the amount of material you need to construct a habitat.
All right, so let's see how this idea works out.
[Malott] There ya go, so that's the dome.
[Montes] But actually, what we wanna do is...
-[Malott] Ah. -[Montes] Turn it over.
[Sima Shahverdi] A vertical habitat?
Like stacking on top of each other?
[Downey] Life on Mars
is not as simple as Bowie made it sound.
The air is a toxic cocktail of mostly carbon dioxide.
There's basically no gravity, no water.
Dust storms can last for months,
from minus 50 to minus 200 degrees Fahrenheit.
[Montes] Probably the biggest challenge
is the temperature.
The habitat has to survive the daily temperature swings,
as if you were traveling
from the equator to Antarctica every day.
They are working there, they are eating there,
they are sleeping there, they are hanging out there.
What kind of a lifestyle do they have?
It should keep you protected from that dangerous outside,
but it shouldn't give you the feeling
that you're in prison.
[Montes] There's a few, we can call them, like,
atmospheric gradients in here,
loud to quiet,
uh, another one could be dirty to clean,
and the third one, dark to light.
[Malott] The A.I.,
by doing lots and lots of iterations,
it discovered a very optimized result.
That concept of this kind of a vertical egg shape
is actually gonna perform a lot better
for a variety of reasons.
[Downey] But Mars isn't the only planet
where A.I.-powered design
is finding new, important applications.
There's an emerging field called geodesign,
which is really about saying how do we take A.I.,
merging that with geography,
merging it with civil and urban planning,
put them together,
and they can say "Well, what happens
if we put the wind turbines here?"
"What erosion is gonna happen
if this pier gets put in in this area?"
"How much density should you be allowed to build in this area?"
"How much water is there?
Is that water polluted?"
Geodesign has a profound ability
to change our lifestyles,
especially around cities and supercities.
[Downey] Generative design helped propel David's group
to the finals of NASA's cosmic competition, in Illinois...
-[Montes] You're the guy? -I'm the guy.
[Downey] ...going up against a team from Penn State.
Three, two, one, go.
[Downey] As much as possible,
everything here should simulate the process on Mars.
They have 30 hours to construct
a one-third scale model of the habitat,
which is about two stories high.
It has to be built
with an A.I.-controlled 3D printer
on its own.
[Malott] We cannot test a building on Mars.
We can't afford
to ship that equipment to Mars in the first place,
'cause it's too heavy,
so how do we know that what we build is strong enough?
And the only way to do that is through A.I.
[Downey] The A.I. uses computer vision cameras
to keep tabs on progress and adjust as needed.
It's the same tech that's used to print stuff
like auto parts and jet turbines.
Well, maybe it'll just kind of do its...
-It might do this. Yeah. -...back and forth.
[Chris Botham] We definitely want
to utilize artificial intelligence
so that the robot or the system
would actually understand what it's doing
and become self-aware of the process.
So we're in auto mode?
Do you have a timer going just in case?
[Malott] You have the robot, and it's very capable,
but it's sort of blind, and it can't think,
it's just going to do what it's been told to do,
so we need to basically take Jeffrey and Chris' experience,
and give that experience to the robot
so that when it goes up to Mars,
it makes the decisions that Jeffrey might've made,
it makes the decisions that Chris might've made,
it might even know that the best decision
is, like, the average of what those two guys think.
I'm gonna time this layer.
[Downey] And remember, there's no Home Depot on Mars,
so A.I.-powered robots
will also need to find, create, and process
all the raw materials themselves.
[Malott] So we're printing with what we call Mars-relevant materials
that we find on Earth, but you can also find on Mars.
People who are really interested in space
actually also really, really care about the Earth...
so the idea of harvesting a building
has always appealed to me.
The material that we selected to build on Mars,
it's renewable, fully recyclable,
and, uh, it's biodegradable,
and as it turns out, that material is super strong.
[Downey] The material
is a mixture of basalt, a common rock on Mars,
and plastic polymer made from corn.
The problem is you can't find corn on Mars.
It's gotta be grown... by robots.
It's going to take a while
before we can grow enough corn to print buildings on Mars,
but we do envision a Mars agriculture.
[Downey] Growing corn on Mars...
is that even possible?
And if it is,
can we figure out a way to grow any kind of food,
It's a climate problem,
but we also gotta figure out the automation part.
[Rothschild] Our ancestors in the Americas
didn't bring everything with them
that they would need for the next thousand years.
They all lived off the land,
and I firmly believe that that is the solution
to colonizing the moon and Mars.
It is extremely expensive to launch anything into space,
and so rather than bringing up all the food and so on,
I believe that what we should do is recreate that scenario
and grow something on site completely done robotically...
and so you need to have some kind of A.I.
being able to make those immediate decisions.
This is science fiction,
but it's also right on that tipping point of innovation
where it's becoming reality.
[Downey] Back on Earth,
robotics is already starting to transform farming.
For now, the work is mostly happening
in a little town in the Netherlands,
but it's not hard to imagine
how far the implications might reach one day.
[Leo Marcelis] I have a passion for plants.
I had it already when I was a very young child.
I grew up on a farm.
I wanted to understand "Why is it growing in this way,
and how can I make the plant grow even in a better way?"
[Downey] Leo's a plant guy...
a plant guy whose research
is not just making farming more sustainable,
but might revolutionize a whole industry.
The area is one compared to 250 of the USA.
[Downey] That's almost twice the size of Manhattan.
Some people call us
the Silicon Valley of agriculture.
[Downey] Turns out, this tiny European country
is now the world's second-largest exporter of fresh food,
thanks to guys like Leo, who are helping to cultivate
some of the most sophisticated and productive greenhouses in the world.
[man speaking Dutch]
[Marcelis] Duijvestijn is a tomato grower.
[man speaking Dutch]
[Marcelis] The greenhouse here
has the size of 50 football fields,
and they are very well known for being extremely high yield
in a very efficient way.
[Downey] And the results are amazing.
These greenhouses produce
seven times more tomatoes per acre
than a traditional farm.
[Marcelis and man speaking Dutch]
The efficiency in a greenhouse
is determined by many different things.
And they give exactly what the plant needs,
not more, not less.
But it is not just the roots,
it's the above-ground environment,
the air humidity, carbon dioxide concentration.
[Downey] Sensors hidden among the plants
generate a constant stream of data,
including temperature, moisture, and soil nutrients.
It's all about engineering climate and optimizing food.
[Marcelis] All that information
comes together in their computer to control the climate.
It's very important to collect the data
and see some trends in it...
...so I can better do my climate control.
You have to look every day?
Every day, I check.
[Marcelis] It's massive data.
You cannot just look at an Excel file and see all those data.
That just does not work,
and that's where the artificial intelligence comes in,
to making use of all those data
to control the crop.
People from all over the world,
they're coming to Wageningen University to study,
because it's ranked number one in agriculture.
[Rick van de Zedde] The robot is taking a leaf off.
We generate a model...
it's not only measuring the leaf,
but also collecting the leaf samples.
[Downey] Despite being on the cutting edge,
Leo and his team are still in the seedling stage
of A.I. farming.
[van de Zedde speaking]
[Marcelis and van de Zedde speaking]
[Downey] To help him innovate further, he's brought in a guy
at the intersection of biology, A.I., and robotics.
But the magic comes in
when we start, like, making sense of the data,
looking at how a batch of plants
can be compared to another batch of plants
where they had different treatments,
and if we understand the differences
between these batches,
we can get a better understanding
of how to optimize plant growth.
[Downey] A.I. is what makes the magic,
by helping them see what they otherwise could not.
Using ultraviolet and infrared light,
Rick and his team are training A.I.
to measure photosynthesis in plants.
The A.I. is really helping us
to bridge the gap between the raw data
and the desired outcome.
[van de Zedde speaking]
[Downey] By using computer vision,
Leo and Rick can measure
the light reflected by the leaves.
This allows them to see how much energy
a plant is generating on a molecular level.
[Erik Brynjolfsson] Some of the leading applications of computer vision
are in things like agriculture and manufacturing.
One of the great opportunities
is that while we can see visible light,
there are many other parts of the electromagnetic spectrum,
like x-rays, infrared, ultraviolet.
It's possible to build sensors for those,
and this opens up a whole space of possibilities
for machines to solve problems that we humans can't.
However, that can only be done when the plants are not too big.
[man 2 speaking]
[Downey] To make things even more efficient and awesome,
Rick's team is working on an automated robotic arm.
[Downey] With the combination of depth perception
and the mobile arm,
the A.I. has the ability
to recognize leaves, stems, and fruit
in the dense greenhouse environment.
So it can see one leaf after the other,
or the fruit behind the leaf.
[van de Zedde] Yeah.
[Marcelis] There's a lot of labor in all those greenhouses
for the harvesting of the fruits.
[van de Zedde] We on purpose selected bell peppers,
because actually that crop is highly complicated.
The bell peppers are much more hidden from view,
so we actually generated a virtual greenhouse...
and we trained an A.I.
to figure out what are the leaves,
what are the stems, and what are the fruits,
and also measure if that particular pepper
is ripe enough to be harvested.
The holy grail is that
there will be greenhouses all over the world
that are run by robots and A.I.
[Downey] If what happens near Amsterdam
does not stay in Amsterdam,
and A.I. precision farming
becomes as commonplace as a tomato or bell pepper,
can you imagine the implications for the human race?
[Marcelis] My dream of the future
is that we can grow the plants in a very efficient way,
where there is not any harm for the environment,
and that there is sufficient food
for people all over the world.
[Downey] Leo's dream can help sustain life on this planet.
But what about elsewhere?
Can we go that far?
Where the climate's more fickle,
and nuts and peppers are hard to come by?
Not to mention synthetic moldable polymers.
I gotta be your eyes.
Nice and easy.
[Malott] We picked two materials that scored well,
which was this biopolymer and this basalt,
which is a composite, which is super strong.
The other team is using concrete.
One of the main differences between polymers and concrete
is that if concrete has to stop,
concrete begins to harden in the pipes,
and you can't un-harden it,
if we had to stop printing for whatever reason,
you can reheat it, and it's back to its original state.
[Botham] You can kind of see the first coursing of that layer.
That's kind of where it made that shelf.
I mean, this is, what, six layers above that?
[Downey] Remember, it's all about adaptation.
Because humans won't be there to supervise,
the A.I. robots need to be able to observe what's happening,
recognize if there's a problem,
come up with a solution to fix it,
and fix it.
-Up a little bit! -Oh!
We want to utilize machine learning in things
so the system adapts to different circumstances.
[Downey] Same goes for this competition.
Because the robots are supposed to operate autonomously,
the teams are penalized
for any interventions they have to make.
[Botham] The two types of interventions that we're dealing with
are remote and physical.
As long as we're pressing buttons
and we're not adjusting the physical system at all,
those are remote interventions.
That doesn't count against us nearly as much.
[Downey] The robots will be alone on Mars
long before humans arrive,
so if Jeff and Chris want the A.I.
to learn to work autonomously,
they need to teach and then step back...
[Downey] ...even if they see mistakes a-comin'.
[Botham] Is it falling in?
[Montes] A little bit.
[Downey] The humans see the problem,
but the robot doesn't.
Because we're printing that layer so quickly,
it's getting a lot of heat buildup,
so it's slumping in that area.
[Montes] Okay, we need to do something about it now.
Let me be your eyes.
[Botham] Just tell me when I'm, like, halfway...
-Okay. -...or three quarters.
Three, two, one, go.
[Earle] Every time Jeff does the countdown,
Chris slows the printer down
to fill that spot in the print a little bit better.
Two, one, back.
[Downey] It's a tough call.
Do nothing, and the whole thing may fall apart.
Intervene, and you get a penalty,
each of which reduces their chances of winning.
[Montes] Three, two, one...
[Shadi Nazarian] Remote, remote!
Remote intervention. Sorry.
[Downey] Jeff's only salvation
is that Penn State is having problems of their own.
[man] It's over there.
I think it's fine, isn't it?
[man] Yeah, I gotta get far away.
[Montes] Three, two, one, return.
[Earle] Is it getting better, Jeff?
We have to keep doing it at a slow speed.
[Shahverdi] So, how many layers are left?
We don't really know.
It's probably around like 55 to 50, something like that.
-Fifty-five? -Yeah, so...
Do you think we can make it?
[Botham] With the race against the clock,
we have to get to a certain diameter up towards the top
so that our skylight can actually sit on top of the form.
If we don't get high enough, it'll just fall inside.
Three, two, one, go.
[Downey] Penn State is almost done.
Finishing faster does not win you the competition,
but it counts for something.
[Nazarian] We need to complete the other cone.
That's what's happening now.
[cheering and clapping]
[Malott] Okay, it looks like
Penn State has just completed their cone.
They are celebrating.
Okay, now we only have ten minutes.
[Downey] Meanwhile, Jeff's team still needs to put a roof over their head.
Uh, we should continue until the last moment.
[Downey] The top must be narrow enough
to support the round skylight
that will complete the enclosure.
[Botham] Um, Jeff?
-[Earle] Jeff? -Yeah?
Are we printing another layer?
-[Montes] Yes. -[Botham] No.
Chris says no.
[Montes] One more.
We can't do one more. It's 5:54, Jeff.
[Montes] One more.
-[Earle] This is the last layer! -Yeah, I know.
Tell me when.
[Downey] While Penn State's done...
-Manual mode? -Yes, please.
[Downey] ...A.I. Space Factory stops the printer
so they'll have the last five minutes
to install the skylight.
[Botham] Relying on you.
[Montes] Yeah, I'll tell you if it's gonna fall through.
[Botham] I can't see it from here, so...
[Montes] I know, I know.
[Downey] If they fail, they're likely to lose the competition.
[Malott] We only have one shot at getting the skylight,
so it's all or nothing.
[Montes] Perfect so far.
-[Earle] Whoa, stop! -No! Stop!
-Down! -[Montes] Down.
[Montes] All right, hover it on top,
and then press "play" in the code.
Down 18 inches.
Nice and easy.
[Botham] Tell me when, tell me when, Jeff.
Go, keep going slow like that.
Keep going. Keep going. Keep going, keep going.
It might fall. It might fall.
-[falling] -[crowd] Oh!
[Malott] The material didn't have enough time to dry.
The weight of the skylight
pulled in on those... those hot layers,
and the skylight... Oh, it just fell.
I heard the thud, and I was like, "ah."
For just that glorious few seconds,
it was there, it was there on the top.
[Downey] The final score
depends on several key structural tests.
[Malott] One of the tests is the smoke test.
NASA is going to take a smoke charge
and put it inside of the printed habitat.
[Downey] The smoke test is intended to prove
the habitats are air-tight...
...but without a sealed skylight in place...
We're gonna be a chimney.
[Downey] With a whopping zero points on that one,
A.I. Space Factory
is comfortably in the... underdog position
heading into NASA's other key evaluation...
...the crush test.
[Downey] Mars is unforgiving.
Needless to say, the structure must be strong and resilient...
to survive the competition, yeah,
but also for life on Mars, because, you know,
they've got those dust storms that last forever
and giant space rocks hurtling at you
at a million miles an hour.
[Montes] I don't even know how heavy that excavator is.
It's the biggest one I've seen in person.
All right, here we go.
[crowd] Oh! Whoo!
Oh, my God!
This is insane.
[Earle] About to crush our habitat.
Is it happening?
Do ya see it lifting?
The CAT's lifting. See it?
[Earle] Oh, shit!
Its entire weight's on that thing.
Oh, my God, that's crazy!
-[Botham] Dude! -[laughter]
[Montes] Dude! Holy shit! That's so much weight!
-Oh, those are the... -[cracking]
-[applause] -[Earle] Yes! That was nuts!
That whole thing was awesome!
-[Shahverdi] Yay! -[Earle] That was so cool!
[Downey] A.I. Space Factory's structure
has resisted the crushing force of a 90-ton excavator.
Will it be strong enough to endure life on a planet
that doesn't have what we might consider
good quality of life?
Time will tell.
For now, it's all about the judges.
Will this surprising result be enough to sway them?
[Montes] I'm stressed. My... my body can feel that I'm stressed.
My back is pretty tense. [laughs]
So we're going to award all our money...
first place goes to
team A.I. Space Factory!
Couldn't have gone better.
[Downey] Life on Mars is not possible...yet.
Getting there, let alone surviving,
is orders of magnitude more difficult than the moon,
but NASA is exploring it...
And now, so are some private companies,
like Elon Musk's SpaceX.
They envision a future
where intergalactic travel is not just possible,
but normal, like taking a Caribbean cruise.
[man] Welcome to space.
[Downey] Some even think it's necessary,
because if we can't fix the planet...
aren't moonshot projections like Mars...
Leo's precision farming...
ultimately about adaptation,
to a changing climate,
a plan B in case we can't get our act together?
Or maybe it's just about a better tomato.
[Marcelis] The population is growing,
the demand for food is rapidly increasing,
so artificial intelligence holds the key
to improve the yield.
We wanna be the first people to build outside of our planet,
and the only way to do that is through A.I.
If we wanna go to Mars,
if we wanna populate other planets,
these types of things require these advanced technologies.
Space technology is the ultimate test
of ideas and ambitions.
I'd like to take some of the DNA from our work
and revolutionize how things are built on Earth.
We should always think about ways
in which what we're doing improves Earth.
[Montes] Hey, I'm gonna do the beam with Riedel.
-Okay. -All right.
[Malott] The load test of our material.
Specifically, it's a bending test.
[man] Are you guys happy with the centering pretty much?
Yeah, it's centered, I think.
[Malott] Twenty-four, 25, so we have to hit 40.
We're at 30, we're at 75% of the way there.
It's deflecting down ever so slightly there.
Okay, we got it, we got it, we got it!
-[applause] -All right.
It passed, at least.
Now to see it shatter.
I dare these guys to try to break it.
[Earle] Now you can see it bending.
[Malott] Yeah, now you can definitely see the bowing.
It... it's gonna pop.
[cheering and laughing]
Oh, man, that's great.
It bounced pretty high.
Is that what it got to?
Must have, yeah.
The amazing part is that our material
doesn't have any steel reinforcement in it.
This is purely that Martian material
made from biopolymer and basalt fiber.
-It's a good one. -[Malott] Wow.