AI’s Most Promising Alien Hunters

A rising generation of well-credentialed scientists is betting that algorithms will turbocharge the search for extraterrestrial life. By Jordan Robertson and Drake Bennett

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:01):
AI's most promising alien hunters. A rising generation of well
credentialed scientists is betting that algorithms will turbocharge the search
for extraterrestrial life by Jordan Robertson and Drake Bennett, read
aloud by Mark Leidorf. When Laura Domine was getting her
pH d in physics at Stanford, her research was on neutrinos,

(00:25):
elementary particles minuscule even to physicists that to lay people
sound made up. Neutrinos are almost massless and electrically neutral,
and therefore they pass through matter as if it were air.
Trillions of the so called ghost particles are zipping through
you right now, unnoticed, continuing journeys that began for many

(00:45):
of them a single second after the Big Bang. The
detectors built to find evidence of neutrinos are themselves fantastical things.
These cavernous chambers deep in the earth are filled with
heavy water or liquid argon and lined with exquisitely sensitive
photosensors or grids of delicate wiring. Every once in a while,

(01:06):
thanks to something called the weak force, a neutrino will
react with a subatomic particle inside the that and be
rendered detectable. By human instruments. Domine worked on a machine
learning algorithm that could spot these reactions, helping physicists continue
to piece together their portrait of the neutrino and, through it,
the workings of our strange universe. After earning her doctorate

(01:29):
in twenty twenty three, Domine didn't join classmates who were
going on to traditional astrophysics research programs or tech companies
or algorithmic finance firms. Instead, she took a postdoctoral fellowship
in Cambridge, Massachusetts, at the Harvard and Smithsonian Center for
Astrophysics and its Galileo Project, which is building a different

(01:50):
kind of detector on a patch of blacktop in the woods,
a half hour outside of Boston. The Galileo project's array
of equipment includes acoustic sensors, radio frequency spectrum analyzer, a
charged particle counter, a weather station with a magnetometer, plus
several cameras, including eight infrared ones. Housed in a twenty

(02:10):
inch wide dome reminiscent of R two D two's head.
This unprepossessing thicket of instruments is pointed at the sky
looking for what are known in the small research community
dedicated to studying them as unidentified anomalous phenomena or UAP.
The rest of us call them UFOs. Domine and her

(02:31):
fellow researchers are willing to consider all explanations for the
anomalies they might uncover, and to them, that means being
open to the possibility that what they'll find are signs
of intelligent extraterrestrial life, not in some distant galaxy, but
in our own terrestrial airspace. Determined to apply mainstream scientific
rigor to their unorthodox quest, the Galileo team has committed

(02:54):
to sifting an entire sky's worth of data twenty four
hours a day. It's something there only able to do
because of recent advances in artificial intelligence. This is the
only way to solve this. Domine says. The availability of
such tools has coincided with a cultural shift, one fed
by a cascade of US Department of Defense disclosures about

(03:16):
its own research into ufology. Although the community of respectable
alien hunting academics like Dominae is still small, it's no
longer confined to the fringes. Research programs similar to Harvard's
have sprouted at Wellesley College, Germany's University of Wurtzberg, and
the Nordic Institute for Theoretical Physics along with the Pentagon.

(03:37):
These programs are often run and staffed by people with
blue chip resumes, such as dominees, but algorithms are proving
to be powerful partners. Rapid improvements in AI software and
the computer servers and other hardware needed to run it
have made it possible to process huge amounts of data
in real time from multiple sources, including cameras and other

(03:59):
instruments deployed in the field. These systems are advancing beyond
their human training, teaching themselves to spot entirely new kinds
of objects in the sky. Nobel laureate Enrico Fermi summed
up the conundrum of extraterrestrial life most succinctly back in
nineteen fifty. Where is everybody? He asked, that is, with

(04:21):
the universe so vast and so rich in planets not
unlike Earth, why haven't we found evidence that we're not alone?
To that end, astronomers use spectroscopes to peer at planets
light years away, assessing whether their atmospheres might contain the
same chemical signatures as ours, evidence potentially of oceans full
of primitive algae like organisms. A recent paper claims to

(04:44):
have found just that. Looking for evidence that aliens are
sending some sort of advanced technology to our planet, however,
is something very different. It's historically been a matter of
examining grainy photographs or nodding solemnly along to witness accounts
of slow eyed humanoids bathed in light, and it's long
been plagued by hoaxes and an openness to a cult thinking.

(05:07):
The public perception of ufology, however, changed on December sixteenth,
twenty seventeen, when a New York Time story revealed that
the Defense Department had itself been secretly studying UAP. The
online version of the story included videos taken from US
Navy fighter jets showing oval blobs that seemed to fly
in physics defying ways. The revelations, after decades of denials

(05:31):
from the US government about its interest in the topic,
prompted Congress to order the Defense Department and the Office
of the Director of National Intelligence to jointly prepare an
annual public report outlining their activities in the area. Congress
also mandated that the Pentagons stand up a new, more
public UAP research program, the All Domain Anomaly Resolution Office.

(05:54):
Sean Kirkpatrick, a materials physicist and longtime defense and intelligence
official was aaro's first director, and he acknowledges some of
the cultural challenges. If NASA and the scientific community are
talking about finding extraterrestrial life, as long as that conversation
is confined to the far reaches of the galaxy in
the universe, that's a very scientific debate, he says. But

(06:18):
as that conversation gets closer and closer to Earth somewhere
around Mars, it starts moving into a more conspiratorial bent.
Once you cross the stratosphere, it's completely out of control.
Domine's own obsession with intelligent extraterrestrial life dates to her childhood.
She grew up in a small apartment in an upscale
western suburb of Paris. Her father, a self taught IT consultant,

(06:41):
immigrated from Vietnam as a teenager, and her French mother,
blessed with an error for languages, worked as a hotel
receptionist in Paris before having Domine in her two siblings.
Domine was an indefatigable reader as a child and a
voracious learner of languages, not only French and English, but
also German and Japanese. She also studied Swahili. She sinsated

(07:03):
others with varying degrees of proficiency, her father's native Vietnamese,
the African Twee language, Cherokee, and the endangered yupik Eskimo language.
I was confident that someday I would be able to
speak most languages that existed in the world. She recalls,
she wasn't so much interested in conversing fluently, though, as
unpacking the patterns of their syntax. They were to her

(07:26):
a series of puzzles. By the time she started high
school two years early, she had taught herself various computer
languages and chunks of physics too. Domine particularly loved the
books of Jacques Valais. The French astronomer is known in
Silicon Valley for doing some of the foundational research that
led to the creation of the Internet everywhere else, though

(07:47):
his prominence comes from a parallel career as a globe
trotting investigator of UFO sidings. The Ufologists played by Francois
Trouffau in Close Encounters of the Third Kind is based
on him. Domine devoured Valet's books about his investigations, as
well as Passport to Magonia, in which he explores the
ways that accounts of modern extraterrestrial encounters mirror the supernatural

(08:11):
elements of much folklore and mythology. Domine's parents shared her
deep enthusiasm for the topic. Her dad had the family
computer set up to participate in SETI at Home, a
global crowdsourced computing project to analyze radio telescope data in
search of messages from deep space. Domine became convinced that
the questions Valet was asking comprised the largest gaping hole

(08:34):
in our understanding of reality, and that studying astrophysics was
the best bet to tackle them. But as she progressed
through France's prestigious Echol Polytechnique, then through graduate school at Stanford,
where she focused on particle physics, it was hard to
see a way to apply her growing expertise to the subject.
In early twenty twenty two, Domine was a year away

(08:56):
from getting her doctorate, after which she planned to leave academia.
She was thinking of working for a climate change startup,
or developing AI tools to help preserve endangered languages, or
taking a year off to become a master woodworker. That April, however,
she read an article about the Galileo project and its founder,
A Vi Lob Lobe, a theoretical astrophysicist, Harvard professor and

(09:18):
one time chair of the university's astronomy department, made his
name with bold ideas about black holes, the Big Bang,
and the birth of stars. Today, though his career is
in its second act. In twenty eighteen, Loeb and a
junior researcher published a paper suggesting that aw Mua Mua,
an unusually shaped space object whose trajectory through our Solar

(09:39):
System had mystified astronomers, might in fact be an alien
space probe. That provocative hypothesis, which Lobe expanded into his
best selling book, Extraterrestrial, The First Sign of Intelligent Life
beyond Earth, made him a celebrity and led to conversations
with a collection of wealthy men who shared his interest
with their money he found at the Galileo Project in

(10:01):
twenty twenty one. In twenty twenty three, Loeb led an
expedition to comb the seafloor near Papua New Guinea for
what he's publicly suggested might be the remnants of another
space probe that had burned up in the atmosphere. He
took along a film crew and documented his progress in
a series of daily medium posts and media interviews. The
universe is a very unusual place, he says in an

(10:24):
interview at his home in Lexington, Massachusetts. You can't be
dogmatic when most of the stuff you discuss is not understood.
Lobe is a deeply controversial figure to his fellow astrophysicists,
who argue that his eagerness to publicly hypothesize about extraterrestrial
technologies goes along with a refusal to consider less splashy

(10:45):
explanations for the various phenomena in question comets, for example,
or mundane meteoroids. But the concept of the Galileo Project
as a place where there were resources and money for
serious UAP research exhilarated Domine when she read at it.
She was particularly intrigued by the concept of the observatory.
You don't have to believe in alien space probes to

(11:07):
see its value. Finding something would be extraordinary, but not
finding anything would be clarifying too. That's how science works.
Domine immediately applied for a postdoc position. A week and
a half later, Loeb called to offer her a job.
We'll be right back with AI's most promising alien hunters.

(11:31):
Welcome back to AI's most promising alien Hunters. When she
got to Cambridge in April of twenty twenty three, Domine
met Richard Clutty, the first postdoc Lobe had hired for
the project. Since then, the two have worked together closely.
Like Domine, Clutty is the first in his immediate family
to get a college degree, though his path was more winding.

(11:53):
The son of a supermarket manager father and a bookkeeper mother,
he grew up in South Africa in various suburbs of
Cape Town and spent his childhood making radios, tesla coils,
plasma globes, microwave guns and other electronic gadgets. He moved
to England in two thousand four when he was twenty
and tended bar for five years before going to the

(12:13):
University of Westminster, where he studied computer science and engineering.
He was only persuaded to get his doctorate by his
now wife, herself a scientist. Before Galileo, he had a
postdoc position in data science at the University of Cambridge.
Clutty also has an old fascination with aliens, though it's
mixed with scorn for what passes for research into the matter.

(12:35):
There's so much nonsense out there. He says, every time
some new information would come out, you'd realize its bunk,
and someone's trying to make money. Even well intentioned UAP
research is a small data affair. Ufologists such as Valet
are detectives rather than scientists, examining fragmentary reports from witnesses

(12:55):
who are themselves often mystified by what they've seen. The
Galileo Project Observatory is meant to rectify that, radically expanding
the amount of data about potential UAP and establishing a
scientific baseline for studying them. The observatory's instruments gather information
from the sky around the clock, but the vast majority
of the time, just as in a neutrino detector, nothing

(13:18):
interesting happens. The plan is for software to do the
onerous work of filtering out all the easily identifiable aerial
phenomena and focusing in on the anomalies, sending them along
for further human analysis by Galileo Project members and the public.
Domine and Klutty are using open source computer vision programs
of the sort found in self driving cars, but the

(13:41):
Galileo Project presents distinct difficulties. The traditional machine learning approach
is to teach a computer to recognize a cat by
showing it lots and lots of cats. But unlike a cat,
a UAP is not a thing defined by particular traits.
It may turn out to be a thing no one
has ever seen before. We don't know what we're looking for,

(14:03):
Clutty says. We don't know anything about its movements. That's
the whole point of the project, to define what a
UAP is. As a result, Dominate and Clutty and the
rest of their team, three other full time observatory staffers,
plus a dozen or so frequent volunteers, are training their
software to recognize all the normal stuff in the skies,

(14:24):
the better to recognize when it's seeing something strange. In effect,
their software models will learn what a cat is by
learning all the things a cat is not. To that end,
Klutty is leading an effort to build massive databases of
real and computer generated images of all the known airborne objects,
a census of the skies. He calls it. The Trove,

(14:46):
draws on aerial images already captured by the observatory. To
supplement that, Clutty uses an open source animation program called Blender.
The software behind the OSCAR winning animated feature flow to
generate one hundreds of thousands of synthetic images of planes, birds, drones, balloons, blimps,
and other objects with planes, for example, or birds that

(15:09):
meant different models and different species at different altitudes, in
different orientations and morphologies. A plane with its landing gear
up and then down, a bird at various points in
its wing stroke. Klutty crammed all the images together in
surreal crowded skyscapes to feed into data analysis software. The
training sessions took place in twelve to twenty four hour

(15:31):
blocks on Harvard's Cannon Compute Cluster, a collection of hundreds
of networked servers and AI friendly graphics processing units spread
over three data centers in the area. Despite the power
of the technologies at their disposal, it can be slow work,
but the Galileo team has been fortunate in the number
of volunteers who've asked to help out an quit Biswass

(15:53):
is a high school junior from Charlotte who contacted Lobe
out of the Blue in March of last year. I
never knew there was various research going on into this,
he says, I had always associated it with this French subculture.
It's this realm where science blends into this pseudo scientific speculation.
For Biswass, who'd won all the state science fairs, he

(16:14):
could find the project was an outlet commensurate with his
ability and energy level. His role on the team was
to process location data from airplane transponders to cross reference
with the observatory's cameras. He worked at it at night
and on weekends when his schoolwork allowed. In general, airplanes
have been one of the easier objects to teach the

(16:34):
software to identify because of the regularity of their speed,
trajectories and turns, as well as the transponder data biswas
worked with. Birds, on the other hand, are difficult and bugs,
fast erratic flyers often close to the camera lens, are
a nightmare. Cludy says. The object detection software his team
is using, called You Only Look Once or YOLO, also

(16:58):
has trouble with clouds, dust, fluttering leaves near the horizon,
and glare from the sun, which YOLO sometimes mistakes for
the flying objects it's supposed to focus on. Our own
visual processing can be susceptible to the same mistakes, of course,
which might explain some of the anomalies human observers have
reported over the years. The goal is to develop similar

(17:19):
software for all of the instruments, then synchronize them so
the entire observatory reacts in real time to what it senses.
That would mean that when one of the sensors, the
all sky camera or the infrared array, or perhaps the
microphone picks up something interesting, a special zoom camera will
swivel toward that patch of sky, and the rest of

(17:39):
the observatory will begin saving the data is recording. Saving
all of the data, all of the time would quickly
overwhelm the team's storage capacity. The system has been continuously
collecting data since twenty twenty three, but it remains in
the so called commissioning stage, where Clutty and Domine and
others are working to ensure that the software models they
are created can do the basic but vital work of

(18:02):
separating signal from noise. In January, Domine published a paper
co authored with Klutty, low Biswas, and others on the
data from the observatory's array of infrared cameras. The YOLO
software was able to identify thirty six percent of the
planes the cameras picked up by the standards of my
PhD neutrino work that's not great, Domine concedes, but she

(18:24):
expects it to improve meaningfully. It's a demonstration of the
level of rigor that we want to stick with for
our studies of UAP, Domine says, and the level we
would expect from our colleagues. She predicts. The Galileo Observatory
should be able to reliably detect outliers coming from all
its sensors within the next year or two and three.
More observatories are in the works in Indiana, Nevada, and Pennsylvania.

(18:50):
The Pentagon is closely following the work at the Harvard
and Smithsonian Center for Astrophysics. Thus far, AAR has examined
more than eighteen hundred reported UA sidings, most of them
from members of the military. Cross referencing information from sidings
with other data sources within the government, such as weather
and flight records, its researchers have determined that hundreds of

(19:12):
cases have straightforward explanations balloons, clouds, drones, etc. Aaro's current director,
John Kozlowski, is a mathematician and engineer whose detailed to
the office from the National Security Agency, where previously he
led research teams in quantum optics and arcane aspects of cryptography.

(19:32):
He comes to the work less from an interest in
aliens than a broader fascination with needles and haystacks. Among
his unclassified research is work on designing a language agnostic
search engine. Kazlowski says that a small number of the
cases his team has examined, about fifty to sixty, are
true anomalies that have stumped the government scientists and engineers

(19:54):
AARO works with. There are interesting cases that I, with
my physics and engineering background and time in the intelligence community,
do not understand, Koslowski told reporters this past November, and
I don't know anybody else who understands them either. Information
about those cases hasn't been disclosed, but Kazlowski says his
office's aim is to eventually declassify as much as possible

(20:17):
about them. As He's careful to emphasize, the term UAP
is by no means just a shorthand for alien spacecraft.
For the Pentagon in particular, the more immediate concern is
that one of our rivals might have secretly developed a
technology far more advanced than ours. The cases that have
been published have revealed the unsuitability of military sensing technology

(20:39):
for aaro's purposes. The now famous Navy jet footage was
taken by the plane's infrared targeting cameras. All these military
sensors are designed to detect and track large, fast moving
objects like missiles, aircraft, and ships, so they can put
a weapon on them, says Kirkpatrick, Kaslowski's predecessor, They're not
built for science observation. That and the classified nature of

(21:03):
their specifications makes the sort of commissioning work that Klutty
and Domine are doing basically impossible. Also, Kirkpatrick points out,
military sensors have important functions that the Pentagon wouldn't want
to compromise. You can't repurpose missile defense radars for chasing
drones because then they're not doing missile defense. To address this,

(21:24):
AARO has collaborated with researchers at the Georgia Tech Research
Institute to develop its own suite of sensors and software
called GREMLIN for Government Radar Multi Spectral Interrogator. Most of
the details about the project remain secret, but the limited
information that's been released shows that the system consists of
equipment very similar to Harvard's Observatory Radar radio antennas and

(21:48):
telescopes gathering visual and infrared images and electromagnetic radiation. Kazlowski
says only that GREMLIN is installed at a national security
site within US borders where there have been multiple u
UAP sidings. Like Harvard's team, AARO is developing custom machine
learning software for UAP detection, and though Kozlowski suggests that

(22:10):
they're further along in the process, they've encountered similar issues.
The undisclosed location where the sensors are being tested is
relatively bug free, Koslowski says, though in previous tests the
AARO computer vision algorithm seem to be confused by grasshoppers.
The Pentagon's effort is in its infancy and may not
survive long enough to answer any of the big questions. Kozlowski,

(22:33):
in an interview at the Pentagon, says he's seen no
evidence that his program will be affected by the drastic
cuts that Trump administration has begun to somewhat haphazardly impose.
As a candidate, Donald Trump promised more government transparency around
the issue. At Harvard, Domine and Klutty's paths are diverging.
They both plan to continue studying UAP, but while Klutty

(22:56):
is extending his postdoc until twenty twenty seven. Domine will
be going somewhere, somewhere else. Over the past few months,
she's watched with alarm as fellow researchers and academics with
international backgrounds have been detained and targeted for deportation by
US authorities. At the same time, she's felt some of
the fallout from the Trump administration's curtailment of federal funding

(23:17):
for scientific research. She'd relied on the National Oceanic and
Atmospheric Administration for some of her data, information the agency
isn't collecting any more. It's become clear to her that,
despite America's growing interest in work like hers, she won't
be able to keep doing it here. All of the
research positions she's applying for are in other countries.