Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
I'm Kai Wright. On the next Notes from
0:02
America, climate change feels more present tense than
0:04
ever, and it is scary. But can you
0:07
see a light at the end of this burning
0:09
hot tunnel? How hope can be part of the
0:11
solution? Listen wherever you get your podcasts. Researchers
0:19
in Michigan used AI to recreate a
0:21
prehistoric land bridge, complete with
0:24
digital wildlife. She
0:26
described the caribou looked like they were
0:28
roller skating because it didn't exactly walk.
0:31
They say it's Thursday, January 25th. But
0:34
we know it's really Science Friday. I'm
0:42
Sci-fi producer Charles Bergquist. In
0:44
this episode, we'll see how archaeologists are
0:46
using AI to track the paths of
0:49
prehistoric caribou to see where
0:51
artifacts from ancient hunters might be located. But
0:54
first, a conversation about Indigenous art
0:57
and its intersection with spaceflight.
0:59
Here's Ira Flatow. The
1:03
patterns woven in textiles can
1:06
tell a powerful story, and
1:08
Sarah Rosalina knows this well. She's
1:10
a multidisciplinary artist who blends
1:13
ancient mediums and Indigenous knowledge
1:15
with data and new technology.
1:18
She's collaborated with NASA JPL, the
1:21
LA County Museum of Art Tech
1:23
Lab, and her work is currently
1:25
featured at the Columbus Museum of
1:28
Art in Columbus, Ohio, until February
1:30
4th. Sci-fi producer and
1:32
host of our podcast, Universe
1:34
of Art, Dee Peterschmidt sat
1:37
down with Rosalina to talk
1:39
about her collaborations with scientists,
1:41
space colonization, and how she
1:43
views technological advancements through an
1:45
Indigenous lens. Here's Dee.
1:47
When Sarah Rosalina thinks about the loom,
1:50
she thinks about computer programming. It's
1:52
an extension of your body being an algorithm.
1:54
Ada Lovelace, who wrote the first algorithm
1:57
design for a computer, said she'd been
1:59
inspired by the card loom developed in the 1800s,
2:01
which used a binary
2:03
punch card to mass produce intricate textile
2:05
designs. In that approach, blending
2:07
old mediums with new tech sums
2:09
up Rosalina's approach to her own art.
2:11
She's an assistant professor of art at
2:13
UC Santa Barbara based in LA, and
2:16
she's of Waraca descent, indigenous people native
2:18
to what is now parts of Mexico
2:20
and the southwestern United States. She
2:22
works in these old art forms, textiles
2:24
and pottery, but uses AI
2:27
and data visualization as part of the
2:29
creative process. It's a way
2:31
to process her feelings about how modern
2:33
society is progressing. We're at this
2:35
point of the technological frontier, and that's
2:37
actually terrifying for a lot of people,
2:39
especially for people from my background and
2:41
my Waraca background. We're living the
2:44
time of climate change, this possession, the
2:46
rise of AI. And I'm
2:48
always interested in anticipating future forms of
2:50
colonization because it's progress for some, but
2:52
it's not for all. Rosalina,
2:55
who's a fourth generation Waraca weaver,
2:57
was taught indigenous textile work in
2:59
part by her grandmother. It's
3:01
something that really made her feel
3:03
sane. I remember she
3:05
so always encouraged me to weave for mental health,
3:07
but it was also good for exercising your mind.
3:10
Rosalina later found herself in the Bay Area around the
3:12
time of the tech boom of the late aughts and
3:14
learned to code. And there was a
3:16
lot of interesting people that I met at that
3:18
time were very similar to me, a lot of
3:21
BIPOC people working in code, but at the same
3:23
time, tech startups really started to rise and displacing
3:25
a lot of the people that I used to
3:27
enjoy hanging out with. Frustrated, she moved
3:30
back to LA and rediscovered
3:32
her love for textile work. I saw
3:34
so many relationships between the code that
3:36
I was writing and actual designs that
3:39
I was weaving that they couldn't help
3:41
but intersect. It was very
3:43
much like an Aha moment. What
3:45
happens when we bring traditional craft
3:47
or indigenous techniques with emerging technology to
3:49
think about current issues that we are
3:52
facing. Digital technologies are always chasing after
3:54
ways that we could simulate our
3:56
reality, which also produces this way that
3:59
we could reinvigorate. The or reality.
4:01
And. Rosalina doesn't just reenvisioned reality
4:03
with herself. She. Up and collaborates
4:05
with scientists to make a right. It's
4:07
a big role as an artist to work
4:09
with scientists and engineers because we see the
4:11
world differently and there's a lot of value
4:13
on that. One of those collaboration
4:16
was with Nasa Jpl in Pasadena
4:18
and Rosalina learn that they had
4:20
a mutual interest play. The.
4:22
Space Agency was experimenting with simulated
4:24
Martian soil also called regolith to
4:26
potentially construct a livable human habitats
4:29
on the Red planet without having
4:31
to transport heavy building materials away
4:33
from earth. So. They were doing
4:35
a lot of research on regular
4:37
simulating played actually build some of
4:39
the first cel then set of
4:41
basically adobe which made me giggle
4:43
because again it's like how much
4:45
space colonization is depended on indigenous
4:47
knowledge even on another planet. Rosalina.
4:50
Also teaches coil pot construction at U C
4:52
Santa Barbara and indigenous method of making ceramics
4:55
that's one of the oldest in the world.
4:57
Coil. Pots look like what they sound
4:59
like coils of clay or layer on
5:01
top of each other until you get
5:03
your vessel as she wanted to update
5:05
that with a techie merson twist. With
5:07
the help of Nasa engineers, she was
5:09
able to make her own version of
5:11
Martian clay. They surf soil analyses from
5:13
two pills rovers like curiosity. She
5:16
muddled the vessels and the computer and
5:18
than treaty printed them using her Merson
5:20
Adobe. The. Resulting sculptures like
5:22
both futuristic an instance of ribbed
5:25
rest and just ceramics take a
5:27
few shapes. The. Mouth of a
5:29
black hole along cylinder that looks like
5:31
it's eating itself. A vaguely spherical shape
5:33
that appears as though it was crushed
5:35
by the forces of gravity. And
5:38
Rosalina passion for pottery even rubbed
5:40
off on Jpl Engineers. are
5:42
actually made a lot of friends
5:44
southern guy and she's surroundings at
5:46
the time which is also really
5:48
interesting to have marcin cartographers who
5:50
guiding the rover's suddenly be interested
5:53
in actually the chemical compounds last
5:55
eight and we would talk for
5:57
hours on and on making clay
5:59
signing native in Los
6:01
Angeles. Located just a few miles
6:03
away from JPL is the Mount Wilson Observatory,
6:06
which Rosalina has also partnered with. It was
6:08
an important observatory in the early 1900s. Edwin
6:11
Hubble used the telescope to prove that the
6:13
universe is expanding, but discoveries
6:16
like that couldn't have been made without
6:18
the help of female computers. Women who
6:20
analyzed the raw data from the telescope
6:22
and performed complex math that made those
6:24
discoveries possible. But when I
6:26
got there, I realized that female
6:28
computers were mostly cropped and edited
6:30
out of the history of that
6:32
observatory. Back then, the images from
6:34
the telescope were exposed onto glass plates,
6:37
which the female computers used to make their
6:39
calculations. And I found textile
6:41
was a unique way to approach it
6:43
because it is a feminist or a
6:45
female-based craft. So to shed
6:47
light on these women's work, Rosalina took those
6:49
plates and digitized them into a lower resolution
6:52
where each pixel would become a bead on
6:54
a tapestry, which she then assembled
6:56
by hand. But not all
6:59
of these tapestries are neat rectangles. Some
7:01
of them distort and fray as the
7:03
beads progress downwards, looking like a starry
7:06
cosmic jellyfish. Rosalina
7:08
hopes her art doesn't just serve as a
7:10
form of protest, but also provides an alternative
7:12
way of interpreting the world around us, one
7:15
that places a much larger
7:17
emphasis on indigenous knowledge. It
7:19
is very important because a lot of
7:21
the current crises that we're facing are
7:24
a crisis of humanity in many ways.
7:26
And I feel like artists really shine
7:28
that light and also can see the
7:30
world differently than what a scientist or
7:32
engineer does. And we can learn quite
7:34
a bit from one another. For
7:36
Science Friday, I'm Dee Petersmith. Thanks,
7:39
Dee. You can check out
7:41
photos of Rosalina's work at sciencefriday.com/textile
7:43
art. And like I said before,
7:46
her art is on view
7:48
at the Columbus Museum of
7:50
Art in Columbus, Ohio until
7:52
February 4th. There was
7:54
a time when people did not question
7:56
the authority of doctors, but HIV and
7:59
AIDS helped change that. People
8:01
were dying. Attention was demanding.
8:03
We literally had to convince
8:06
the government that there were women
8:08
getting HIV. It was hard to
8:10
fight and it was activists.
8:13
We changed the world. Join
8:15
us for Blind Spot, The Plague and
8:18
the Shadows, a series from the History
8:20
Channel and WNYC Studios. Listen wherever you
8:22
get podcasts. Artificial
8:28
Intelligence is great at detecting
8:30
patterns, which means its calculations
8:32
can help predict the future.
8:35
But AI can also be used to take
8:37
a look back into the past. That's
8:40
exactly what one research team in Michigan
8:42
is doing. Using AI
8:44
to track the paths of
8:46
prehistoric caribou. Why? To
8:49
see where artifacts from ancient hunters
8:52
may be located. Joining me
8:54
to talk about this is my guest,
8:56
Morgan Springer, editor of the Points North
8:58
podcast at Interloc & Public Radio in
9:00
Interloc in Michigan. Welcome to
9:02
Science Friday. Thank you so much for having me.
9:05
Help me imagine what we're talking
9:07
about here before we get to
9:09
the AI caribou. Where is
9:11
this land bridge that researchers are so
9:14
interested in? Yeah, so it's at
9:16
the bottom of Lake Huron, which for listeners that
9:18
don't know, it's one of the Great Lakes. It's
9:20
on the east side of Michigan and
9:23
the official name of the land
9:25
bridge is the Alpena-Amberly Ridge. And
9:27
it goes from northern Michigan to
9:29
southern Ontario, kind of cutting the lake
9:31
at a diagonal. And what's
9:33
the significance of this bridge? Yeah,
9:36
so what I'm going to say, it's going to sound
9:38
obvious once I say it, but the
9:40
Great Lakes didn't always look the way that
9:42
they do now. If we go back to
9:45
the ice age, the glaciers
9:47
are receding. And about 10,000 years
9:50
ago, lake levels were lower than they
9:52
are today. So that
9:54
means land that's now underwater, it was
9:56
above water then, including this ridge, this
9:59
land bridge. And it
10:01
was continuous. It was this causeway
10:03
where people and animals could move
10:05
and migrate back and forth and
10:07
leave artifacts, presumably. And so then
10:10
the water levels rise, it comes
10:12
up, and these artifacts
10:14
are submerged and remarkably preserved
10:16
and protected from development. And
10:19
you know, archaeologists were skeptical
10:21
that they'd find anything, that
10:24
this was going to be an opportunity to find
10:26
artifacts, but they wanted to look anyway. And
10:29
let's get into the details of this. Why
10:31
would someone want to research how
10:34
animals crossed this long-gone path? Yeah,
10:36
so John O'Shea, he's an anthropological
10:38
archaeologist and he's based at the
10:40
University of Michigan, and he wanted
10:43
to find something. And so basically he came up
10:45
with an idea for something he thought
10:48
he could find, something that would have
10:50
survived being inundated with water about 9,000
10:52
years ago. And so
10:54
one of the things they knew about that time
10:56
period was that caribou were the main source of
10:58
food. And they also
11:01
knew that prehistoric hunters made these
11:03
really cool hunting structures, they're
11:06
called drive lanes, and they would guide
11:08
the caribou to these kill
11:10
sites. And so John
11:12
O'Shea, his collaborator, they thought if they
11:14
were made of stone back then, maybe
11:17
we could find them underwater. So it's
11:19
all these hypotheticals, but it
11:22
helps to know where the caribou would
11:24
go so that they can know where
11:26
to look for sites. So
11:29
the idea is if the caribou follow a
11:31
certain path, then humans probably aren't
11:33
far behind, and then it's the
11:35
humans who are leaving these artifacts. Exactly.
11:38
And why couldn't researchers just find
11:41
these artifacts the old-fashioned
11:43
way? Yeah, so technically they did
11:45
find the first one the old-fashioned
11:48
way, kind of. I mean,
11:50
they used side-scan sonar. I think they had
11:52
an underwater robot at the time, but there
11:54
wasn't any AI. But
11:56
regardless, the challenge was that Lake Huron
11:58
is... huge and even
12:00
though the land bridge offers this concentrated
12:04
place, this corridor to look,
12:06
it's still really long. It's about 90 miles
12:09
long. It's about nine miles
12:11
wide and then on top of that you got to
12:13
go 100 feet underwater. And
12:16
here's John O'Shea talking about
12:19
this process. Underwater research
12:21
is always like a needle in a haystack. So
12:24
any clues you can get that help you
12:26
narrow down and focus the kind of places
12:28
you might look at is a real help
12:30
to us. And you
12:32
know, John happened to know the
12:35
premier, one of the premier
12:37
people doing archaeological computer
12:39
simulation. His name's Bob Reynolds.
12:41
He's based at Wayne State
12:43
University. And so their
12:45
idea was they'll create a
12:47
computer model of the land bridge and
12:50
then use AI to help predict sites. And
12:54
how does this AI actually work? What
12:56
kind of information were the researchers plugging
12:58
into the model? Great question.
13:00
Okay, so the first step is you've got
13:02
to actually build the virtual
13:04
land bridge, the Alpina-Amberley Ridge, and
13:07
they use the actual topography. And
13:09
then they start populating it with
13:11
digital caribou. And that's the piece
13:14
that has the artificial intelligence. So
13:16
they create these caribou and they
13:18
give them instructions, their computer algorithms,
13:21
and the instructions basically tell them how to
13:23
behave. A really simple one
13:25
is caribou walk. Okay,
13:28
so the caribou start walking. And then
13:30
another simple one is be aware of
13:32
obstacles and move around them like don't
13:34
bump into rocks or each other. Another
13:36
one is move in groups and break
13:39
apart. And they just they keep refining
13:41
and refining it until the caribou start
13:43
behaving more and more like real
13:45
caribou. And what did
13:47
it look like to watch the
13:49
AI model in action? I'm picturing
13:51
a sort of animated video of
13:53
caribou walking around. But is
13:55
that what the early models really looked like? They
13:58
had some glitches. One
14:03
of the researchers I talked to, she
14:05
described that the caribou looked
14:07
like they were roller skating
14:09
because they didn't exactly walk.
14:11
But they've kept developing it
14:13
and that's really a whole
14:15
other story. Now they have
14:17
an amazing virtual reality where
14:19
they really look like caribou.
14:23
But another glitch was Bob Reynolds,
14:25
the computer scientist at Wayne State
14:27
that I mentioned. He
14:30
talked about this other one glitch that was
14:32
funny. Literally the first model. We
14:34
let the herd run across the land bridge
14:37
and they did not have edge perception and
14:39
so they kept dropping off the size of
14:41
the bridge like lemmings. Oh
14:43
no! RIP to the
14:46
AI caribou I guess. I
14:49
know. And so that's a perfect example of
14:51
where they have to introduce a new algorithm
14:53
and they basically give the caribou a new
14:55
instruction which says, hey you gotta perceive
14:57
edges. So it's a lot of
14:59
trial and error and refining. And
15:02
how well is the AI working today?
15:04
I mean, have the researchers actually found
15:06
any real world evidence based on these
15:08
computer generated paths? Yes,
15:11
absolutely. They've found
15:14
prehistoric hunting sites, they've found
15:16
artifacts. Ashley Lemke,
15:18
she's an anthropological archaeologist also
15:20
on the team and she's
15:22
currently a professor at the
15:24
University of Wisconsin-Milwaukee and
15:27
here she is. We
15:29
could ask this archaeologist, how did you find a site? Or
15:31
how did you know where to dig? And
15:33
for me I can be like, oh well artificial
15:35
intelligence told me. So how
15:38
it works is the caribou developed these optimal
15:40
routes over time. They're going back and forth
15:42
and back and forth. And
15:44
there were a few spots that they went nearly
15:46
every time and they called these choke points. And
15:50
so this is a really obvious place for archaeologists
15:52
to go and look. And this
15:54
is just one example of how AI
15:56
helped. But this one particular choke
15:58
point led them to the site, they
16:00
call it Drop 45. It's
16:02
the most complex hunting structure found in
16:05
the Great Lakes to date. And
16:07
there's a number of things there. There's a
16:09
line of stones guiding caribou to a kill site.
16:12
There was a fireplace with burnt
16:14
earth, incredible 9,000 years old. And
16:17
then there were also these really
16:19
unusual small tools that were unprecedented
16:21
for the season. And you know,
16:24
that's AI. AI helped them find
16:26
that and it saved them time
16:28
and money. And
16:30
now that these amazing artifacts
16:33
are being, little and large, I guess
16:35
artifacts are being found, what's
16:38
next for the team? Yes.
16:40
So keep looking.
16:43
Yes, they found some
16:45
artifacts, but they've just
16:48
scratched the surface. And with
16:51
what they found, they've started to build an
16:53
understanding about what the environment might have looked
16:55
like and how people might have lived. But
16:58
they've also found some totally new
17:00
and, as I mentioned, unprecedented artifacts.
17:02
Here's Ashley Lemke again. None
17:05
of this matches the models we had
17:07
about peoples in this region, which is really, you know,
17:10
it's really fascinating because then you have to go back
17:12
and be like, all right, well, now we have this
17:14
new data, you know, what does that mean for
17:16
what we thought about peoples that were living in
17:18
the Great Lakes? You know, you kind of have to rewrite the
17:20
story. So they keep looking, they
17:23
keep researching and they keep rewriting
17:25
the story. And now that
17:28
it's been proven that this sort of
17:30
application for AI works, do you think
17:32
it'll gain traction in the larger scientific
17:34
community? You know, I don't
17:36
know. I think it should for
17:38
sure. But Bob
17:41
Reynolds, he's the main computer scientist. This
17:43
is not the only project he's worked
17:45
on. So it's definitely something he's working
17:48
on with other archaeologists,
17:50
other scientists. But it
17:52
requires a lot of strong collaboration
17:54
between completely different fields. And
17:57
I know that Bob specifically, he
17:59
leans heavily on students at Wayne
18:01
State to really help make the
18:03
simulation and the virtual reality come
18:06
to life. That's all the time
18:08
we have for now. I'd like to
18:10
thank my guest, Morgan Springer, editor of
18:12
the Points North podcast at Interlochen Public
18:14
Radio in Interlochen, Michigan. Thank you for
18:17
joining me. Thank you so
18:19
much, Sophie. That's it
18:21
for today. Lots of folks helped make
18:23
the show, including Ariel Zich,
18:26
Jordan Smudgick, Diana Plasker,
18:28
and many more. Tomorrow, we'll check
18:31
in on the top stories from the week
18:33
in science. I'm Sci-fi producer Charles Bergquist. Thanks
18:35
for listening. We'll see you soon.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More