[Music]
we’ve all been there you’re walking
around during the night coming from a
party when suddenly you look up and you
notice how wonderful clear the night sky
is maybe some of you even remember an
anecdote from a teacher or relative
telling you that the light from these
stars has often traveled many many years
before it found its way on through our
eyes or it could be that you just
remembered from Carol’s talk like two
hours ago I mean whatever works because
stars are so distant studying them can
be a very challenging task it’s one of
the reasons why I studied physics and
astronomy because how could we learn so
much when there’s so little to work with
astronomers don’t have a control lab
environment where they can carefully
monitor all aspects of their object of
study instead the galaxies are often in
a very chaotic and turbulent environment
and because they’re so far away it’s not
like we can physically go there to get
some samples on top of that we’re stuck
looking at these things from one side
where we basically only get a
two-dimensional projection of something
that’s intrinsically three dimensions
it’s like you’re trying to recreate an
entire house when the only thing you get
is a picture from one side so how do we
even start solving that problem
luckily lights can carry a lot of
information just look at the world
around you and notice how much we can
learn just by looking at things the
color the intensity the way objects
reflect light can tell us whether
they’re smooth or coarse whether they’re
hot or cold whether they’re made from
metal or wood in a similar way
astronomers can extract information from
light coming from stars to learn more
about their temperature their movements
their composition but because we have
vast amounts of data of bill
of galaxies there is no way that we can
do all of this manually and we need
artificial intelligence programs to help
us understand this data for us and it’s
something that they’re they’re really
good at I mean and it’s really not the
boring part of the job to see an
algorithm that’s self learning that you
can see picking up new rules in front of
your eyes can be very self fulfilling
and rewarding it’s a bit like becoming a
parent every single time and without the
horror of actual child births it’s not
like these self probing self learning
programs grow on trees unfortunately
although maybe they do a lot of these
programs founder nature have actually
their root in nature like for example
let’s have a look at how a giraffe is
just that so well to its environment
with the taller neck and legs over the
years there has been a selection where
giraffes that were not tall enough would
be stuck eating leaves closer to the
ground therefore unable to reach this
large pool of nutrition that the taller
ones were able to get to because of this
disadvantage over time they disappeared
because their chance of surviving and
creating offspring are smaller so it’s
because of this harsh filtering where we
add random mutations to introduce new
characteristics that the giraffe as an
animal has become how we know it with
the large neck and legs it’s this
systematically favoring the fittest and
only using those to create offspring
that creates such a powerful mechanism
which we can translate into a program we
could translate it into an algorithm
which is essentially a set of procedures
and instructions a computer can
understand the genetic algorithm is an
algorithm that captures the essence of
evolution theory it does so by creating
a lot of random
models and adding a lot of randomization
and then filtering out those that are
best fitted to solve a given problem to
see how this is used in a real-life
example let’s have a look at this the
top image is a real observed galaxy that
we see in two dimensions the bottom
image is a genetic algorithm trying to
recreate that image and we’ll see it
starts off quite randomly making wild
guesses but over time you can see it
starts picking up new rules and more
information out of this image until it
ends up with something that’s actually
pretty close to the original one
and I can hear a lot of you people think
like okay yet great stuff but if you
just wanted to kind of recreate that
image you could have just copy pasted it
right so why bother the difference is
that from the lower image we have the
exact three dimensional distribution so
we know where the stars are how they’re
distributed how old they are how the
dust is distributed so in the analogy of
the house that we made before it’s like
we’re creating a lot of random houses
I’m taking pictures from all sides until
we have that one picture that we were
looking for genetic algorithms help us
quickly find those that are up for the
case and remove the ones that are
clearly not doing a good job but we can
even do better than this our human eyes
only see a limited amount of colours and
wavelengths but these galaxies emit on
way more than just this
they emit from the very short x-rays
that the doctor for example uses to
check whether your bones or fractures
all the way up to the very long radio
waves and they can take all of his
information together to scan this object
in a much smarter way and extract
information from data that no human eye
has ever directly seen but genetic
elements are not the only
nature-inspired program swarm
intelligence is the intelligence we see
large groups of insects and birds that
seemingly move as one large organism
they do so without ever colliding
without ever communicating and without
even having a leader that orchestrates
the entire scenery swarm intelligence
will be used a lot to solve optimization
problems guide drones and manipulate
satellites for years to come but it’s
not just science fiction swarm
intelligence already solves quite a lot
of problems today with a lot of GPS
systems having algorithms that are based
on ant colonies because hands have
perfected the art of finding the optimal
routes between their colony and a source
of nutrition they do so by first
spreading out randomly until they found
some foods at which point the ant that
found the food starts spreading out some
pheromones these pheromones are then
picked up by the other ants nearby who
are now more likely to follow that path
than the random path before so over time
more and more ants start following that
bad and even reinforce it by spreading
out pheromones themselves so ants have
optimized their travel time in a very
clever way and they can do so very
flexibly very flexible in whenever
sudden changes in the environment take
place but there’s an even more common
example water has the remarkable feature
that it always finds the pad of least
resistance it never runs uphill so by
carefully simulating these water drops
you can create a very powerful mechanism
that can be used to solve optimization
problems and potentially our future
traffic jams so it seems that wherever
we look in nature we always find some
sort of intelligence and this should not
come as a total surprise because a lot
of these laws and mechanisms have
existed for millions of years and have
been fine-tuned to perfect
and we as humans should learn how to
exploit this knowledge and take nature’s
expertise on this if you want to advance
and learn more about our universe as we
speak
artificial intelligence help us
understand how galaxies are formed and
evolved from the Big Bang up until today
they extract information out of these
vasts amounts of data and learn more
about our universe just as we do their
goal being the same as ours to watch and
learn and while we still have quite a
few problems to solve up in the night
sky I bet that there might still be a
few solutions laying right here at our
feet
[Music]
you
we’ve all been there you’re walking
around during the night coming from a
party when suddenly you look up and you
notice how wonderful clear the night sky
is maybe some of you even remember an
anecdote from a teacher or relative
telling you that the light from these
stars has often traveled many many years
before it found its way on through our
eyes or it could be that you just
remembered from Carol’s talk like two
hours ago I mean whatever works because
stars are so distant studying them can
be a very challenging task it’s one of
the reasons why I studied physics and
astronomy because how could we learn so
much when there’s so little to work with
astronomers don’t have a control lab
environment where they can carefully
monitor all aspects of their object of
study instead the galaxies are often in
a very chaotic and turbulent environment
and because they’re so far away it’s not
like we can physically go there to get
some samples on top of that we’re stuck
looking at these things from one side
where we basically only get a
two-dimensional projection of something
that’s intrinsically three dimensions
it’s like you’re trying to recreate an
entire house when the only thing you get
is a picture from one side so how do we
even start solving that problem
luckily lights can carry a lot of
information just look at the world
around you and notice how much we can
learn just by looking at things the
color the intensity the way objects
reflect light can tell us whether
they’re smooth or coarse whether they’re
hot or cold whether they’re made from
metal or wood in a similar way
astronomers can extract information from
light coming from stars to learn more
about their temperature their movements
their composition but because we have
vast amounts of data of bill
of galaxies there is no way that we can
do all of this manually and we need
artificial intelligence programs to help
us understand this data for us and it’s
something that they’re they’re really
good at I mean and it’s really not the
boring part of the job to see an
algorithm that’s self learning that you
can see picking up new rules in front of
your eyes can be very self fulfilling
and rewarding it’s a bit like becoming a
parent every single time and without the
horror of actual child births it’s not
like these self probing self learning
programs grow on trees unfortunately
although maybe they do a lot of these
programs founder nature have actually
their root in nature like for example
let’s have a look at how a giraffe is
just that so well to its environment
with the taller neck and legs over the
years there has been a selection where
giraffes that were not tall enough would
be stuck eating leaves closer to the
ground therefore unable to reach this
large pool of nutrition that the taller
ones were able to get to because of this
disadvantage over time they disappeared
because their chance of surviving and
creating offspring are smaller so it’s
because of this harsh filtering where we
add random mutations to introduce new
characteristics that the giraffe as an
animal has become how we know it with
the large neck and legs it’s this
systematically favoring the fittest and
only using those to create offspring
that creates such a powerful mechanism
which we can translate into a program we
could translate it into an algorithm
which is essentially a set of procedures
and instructions a computer can
understand the genetic algorithm is an
algorithm that captures the essence of
evolution theory it does so by creating
a lot of random
models and adding a lot of randomization
and then filtering out those that are
best fitted to solve a given problem to
see how this is used in a real-life
example let’s have a look at this the
top image is a real observed galaxy that
we see in two dimensions the bottom
image is a genetic algorithm trying to
recreate that image and we’ll see it
starts off quite randomly making wild
guesses but over time you can see it
starts picking up new rules and more
information out of this image until it
ends up with something that’s actually
pretty close to the original one
and I can hear a lot of you people think
like okay yet great stuff but if you
just wanted to kind of recreate that
image you could have just copy pasted it
right so why bother the difference is
that from the lower image we have the
exact three dimensional distribution so
we know where the stars are how they’re
distributed how old they are how the
dust is distributed so in the analogy of
the house that we made before it’s like
we’re creating a lot of random houses
I’m taking pictures from all sides until
we have that one picture that we were
looking for genetic algorithms help us
quickly find those that are up for the
case and remove the ones that are
clearly not doing a good job but we can
even do better than this our human eyes
only see a limited amount of colours and
wavelengths but these galaxies emit on
way more than just this
they emit from the very short x-rays
that the doctor for example uses to
check whether your bones or fractures
all the way up to the very long radio
waves and they can take all of his
information together to scan this object
in a much smarter way and extract
information from data that no human eye
has ever directly seen but genetic
elements are not the only
nature-inspired program swarm
intelligence is the intelligence we see
large groups of insects and birds that
seemingly move as one large organism
they do so without ever colliding
without ever communicating and without
even having a leader that orchestrates
the entire scenery swarm intelligence
will be used a lot to solve optimization
problems guide drones and manipulate
satellites for years to come but it’s
not just science fiction swarm
intelligence already solves quite a lot
of problems today with a lot of GPS
systems having algorithms that are based
on ant colonies because hands have
perfected the art of finding the optimal
routes between their colony and a source
of nutrition they do so by first
spreading out randomly until they found
some foods at which point the ant that
found the food starts spreading out some
pheromones these pheromones are then
picked up by the other ants nearby who
are now more likely to follow that path
than the random path before so over time
more and more ants start following that
bad and even reinforce it by spreading
out pheromones themselves so ants have
optimized their travel time in a very
clever way and they can do so very
flexibly very flexible in whenever
sudden changes in the environment take
place but there’s an even more common
example water has the remarkable feature
that it always finds the pad of least
resistance it never runs uphill so by
carefully simulating these water drops
you can create a very powerful mechanism
that can be used to solve optimization
problems and potentially our future
traffic jams so it seems that wherever
we look in nature we always find some
sort of intelligence and this should not
come as a total surprise because a lot
of these laws and mechanisms have
existed for millions of years and have
been fine-tuned to perfect
and we as humans should learn how to
exploit this knowledge and take nature’s
expertise on this if you want to advance
and learn more about our universe as we
speak
artificial intelligence help us
understand how galaxies are formed and
evolved from the Big Bang up until today
they extract information out of these
vasts amounts of data and learn more
about our universe just as we do their
goal being the same as ours to watch and
learn and while we still have quite a
few problems to solve up in the night
sky I bet that there might still be a
few solutions laying right here at our
feet
[Music]
you
Please follow and like us: