Inteligencia artificial y la materia oscura del universo
Translator: Gisela Giardino
Reviewer: Sebastian Betti
When I was a girl and for several years
my dad used to take me to visit
his friend Manuel on Saturdays.
Manuel Sadosky was a wise person
we both greatly admired.
I still remember the passion in his eyes
when he talked about anything.
From scientific policy
to his Clementine stories.
Clementine was the first computer
Argentina had
that was used for scientific purposes.
It occupied a whole room and he himself
had brought it to the country in 1960.
In the living room of his tiny flat
filled with books
I listened to the questions he asked me.
If I knew the origin of zero, of infinity,
and many other questions
I had never imagined
they were even thinkable.
When Manuel talked about history
I would leave wanting to be a historian.
Other times, when he talked
about mathematics or philosophy
I’d leave wanting to be
a mathematician or a philosopher.
Those afternoons at Manuel’s
made me feel elated.
I remember I would leave
feeling more curious,
with the impatience
of those who don’t know
and still have a whole world to know.
Like a girl before a new jigsaw puzzle,
who can’t wait
to start putting it together.
Manuel opened the doors
of curiosity for me
and taught me the value of questions.
Over the years I realized
that this is the feeling
you have when you do science.
The more you know, the more doubts you
have and the more you want to know.
After graduating from physics
at the University of Buenos Aires,
I finally found in Cosmology
the space to put together
my passion for mathematics,
physics and philosophy.
And that’s what I do now.
Cosmologists are like archaeologists,
we study the past
to be able to better understand
our present and our future.
Just like archaeologists use ruins
we cosmologists use
the light that reaches us
from somewhere in space.
The light that comes to us,
for example, from a star.
It can take a long time,
millions of years, to reach us.
And when it does, it lets us see
what the star was like in the past.
The farther away the star is
it takes longer to reach us
and farther back in time we can see.
In other words, studying
the distant universe is like time travel.
Back in 1965,
two Bell Labs astronomers,
Arno Penzias and Bob Wilson,
were working with an antenna
they had built to detect
radio waves that came from space.
To be able to measure those dim waves
they had to get rid of every possible
interference in their detectors.
And they did it by cooling them down
at very low temperatures, -269 °C.
But in doing so, they found
a signal they could not explain.
They tried to get rid of this signal
in every way possible.
But they couldn’t.
They even climbed the antenna
thinking that the problem
could be pigeon poo to no avail.
The signal was still present.
The oldest light we can observe
travels towards us from the Big Bang
14 billion years ago.
At first, protons, electrons and photons
formed a dense soup of particles
to very high energies.
This soup oscillated, gravity
made the particles come together,
and photons put pressure outwards.
I like to think of these fluctuations
like a cosmic symphony
where oscillations
are like the notes of an instrument.
Meanwhile, the universe
was expanding and cooling down.
400,000 years after the Big Bang
the first stable atoms formed.
And since then, light travels towards us
on a lonely journey
bringing information
of what was happening at the time.
This light is known as
“cosmic background radiation”.
And the temperature of this light
comes to us with small fluctuations
in different parts of the universe.
The regions with higher temperatures
had more matter.
And regions with a little more matter
grew, because of gravity,
and ended up forming the stars
and the galaxies we observe today.
While Penzias and Wilson
were still trying to fix the antenna
they found out that a group
at Princeton University,
not far from where they were
was writing an article
in which they proposed existence
of this cosmic background radiation.
They immediately realized
the importance of what they had detected.
What they thought was
pigeon poo, was no less
than cosmic background radiation.
The glimpse that reaches us
from the Big Bang.
A few years later, back in the 70s
an astronomer, Vera Rubin,
was studying the rotation speed
of stars in different galaxies.
Her measurements gave results
quite different than expected.
It was expected to see
that the stars were slower
the further away they would be
from the center of the galaxy,
just like with the planets
spinning around the sun.
But she saw that speed
remained constant
even for distances
far from the center.
And for this to be possible
there should be in space
more matter than we can observe.
This was the first sound evidence
of what is known as “dark matter”.
And this pioneering work by Vera Rubin
served to explain lots of observations,
including the cosmic symphony
of the Big Bang I told you about before.
Believe it or not, 85 percent
of the matter all over the universe
is dark matter.
85 percent and still today
we don’t understand what it is.
Unfortunately, Vera Rubin
died a few years ago,
without receiving, in my opinion,
well-deserved recognition.
When I met her at an event
for women scientists 11 years ago,
at the University of Chicago,
there was already
a timid talk around the issue
of gender disparity in sciences.
This was a problem Vera knew too well.
With my research group at Harvard
we try to understand what dark matter is
through its effects
in our observations of the cosmos.
And because we can’t see it directly,
we’re trying to find
other ways to detect it.
For example, we make use of the fact
that its mass deforms space-time.
And that’s why the light
coming to us from any galaxy,
instead of traveling
in a straight line,
deviates creating curves in the sky.
This phenomenon predicted by Einstein
in his theory of general relativity
is known as “gravitational lens effect”.
We look for small clumps of dark matter
because we believe that
we can find clues in them
about the nature of dark matter.
These light clumps
create small fluctuations
in the arches we observe in the sky.
The traditional method of detecting
these little disturbances
is to analyze image by image,
which is a really time-consuming job.
Researchers spend months
analyzing these images
and in general, they don’t find anything.
Until now only two detections
had been made with this method.
Now we have dozens of images
of arches to analyze.
But soon with data
from new telescopes
we’ll have tens of thousands
of these images.
And this kind of analysis
will become impossible at this pace.
We were thinking about this problem
with my group
when we thought of using a new branch
of artificial intelligence:
machine learning.
It aims to develop algorithms
that make computers be able to learn
by pattern recognition.
These algorithms are the ones
that Google translator use
or YouTube when
they recommend you a video
based on your personal taste.
We use it to research
on dark matter.
We train these programs
using hundreds of thousand simulations
of galaxy images.
And we saw that
with these kinds of methods
we can find clumps of matter,
even very small clumps.
I still remember my surprise
this February
when two members of my group,
Ana Díaz Rivero and Bryan Ostdiek,
came to my office
with their first results
and I told them to check them again
because they were
too good to be real.
After many comings and goings
we observed that, actually,
our method works much better
than traditional analyses.
We can detect
clumps of dark matter,
even clumps we once thought
impossible to detect,
and we can do it
in less than a second,
instead of months of researchers
analyzing image after image.
A few months ago, with colleagues
from Harvard University, MIT,
and other universities in the region
we opened an institute
of artificial intelligence
where physicists from different
disciplines can use these methods.
I sometimes wonder what Manuel would say
if he could see what we’re achieving
with these intelligent machines.
And I also think about how our knowledge
about the universe changed,
from Clementine’s time until now.
The large amount of data
we managed to collect
and the more to come
will help us fit in better
our puzzle pieces.
And as the pieces come closer
new questions arise.
And this for me is to do science:
the constant quest
not only for answers,
but also as importantly,
of the right question.
And it doesn’t matter that the pieces
of the puzzle don’t fit perfectly.
They probably never will.
After all, this is part
of the beauty of doing science.