We all like to learn things.
Some of us like to learn a language,
others learn to play musical instruments and others a new sport.
And there are things nobody likes to learn,
but we would love to know.
For example, I would love to know how to play the guitar,
but I don’t want to take any lessons.
Maybe learning Chinese, just because.
Frankly, who wouldn’t like to learn
anything they wanted with the least possible effort?
I need to curb your enthusiasm a bit, because I don’t have a magic recipe.
However, what we know about the brain
is changing at an incredible speed.
We still lack enough to be able to learn
the keys to learning, but today I want to tell you
how are we progressing in this regard.
I studied electronic engineering
and what interested me most, was the area of artificial intelligence
and how to create smart machines that can learn things by themselves.
Then I became interested in how humans,
and animals learn.
There I went to study neurosciences and I learned to measure neuronal activity
and train laboratory animals: mice, monkeys and students.
(Laughter)
Now, the task of a neuroscientist is not that simple.
In fact, the brain is the most complex and intelligent device that we know.
It’s composed of 80 billion neurons, like one of these.
To make it worse, each neuron receives on average,
up to 10,000 connections from other neurons.
This is as if each one of you could listen at the same time
and be influenced by all the people present today at TEDxRíodelaPlata.
When we learn something, our brains change
and they change physically.
Although you don’t realize it, your brain is changing right now.
What does this mean?
It means that our brain,
generates new connections as we learn.
There are neurons that start to connect with other neurons.
And some of these connections
get stronger or become weaker.
How can we study what happens in the brain
when we learn something new?
Well, we need a way to be able to measure neuronal activity
very precisely.
But the most difficult is that we have to keep track
of exactly the same neurons every day throughout the learning process.
This is what we couldn’t do until some time ago.
A new technology allows us to do this,
it’s called two-photon microscopy
and it has opened our scope so that we can study
what happens in the brain when it learns.
In this video, I am showing you neurons responding.
These spikes are active neurons.
Those flashes of light, when they are more intense,
indicate a more active neuron, responding better in every turn.
A lot of my work as a neuroscientist
is about trying to relate the intensity of each neuron
with the behavior of an animal.
These microscopes allow us to follow these same neurons I’m showing here
for weeks.
The same, the same.
This opened the door to studying
what happens in a brain for the first time.
And how can we do this?
How can we study how an animal learns?
Well, we have to get a lab animal
and teach them something they didn’t know
while, at the same time, we measure their neurons
to see how they change as they learn.
Well, I’m going to tell you how we did this in the laboratory.
We take a mouse and we teach them
to detect the position of an object using their whiskers.
You might ask: why the whiskers?
Well, mice are night animals
and, besides, they don’t see very well,
so they use whiskers to navigate their surroundings.
For them, whiskers are like our own fingers.
Here in the video I’m going to show you how a mouse solves this task.
That little black stick you see is the object that they have to detect.
And the mouse starts moving the whiskers like crazy trying to locate it.
The object may appear in different positions.
Now, what we need is the mouse to tell us
where the object was.
How could we teach them that?
Well, we teach them to use their tongue
to tell us if they had detected it or not.
The experiment was like this:
the object could be in two positions,
near the face or away from the face.
If it was close, they had to stick their tongue out
and if it was far away, they had to keep their tongue inside.
And all this was solved in complete darkness.
The craziest of all is that we didn’t give the mouse
any particular instruction.
The only thing we gave him is a reward if it was done correctly.
And little by little, the mouse could learn.
Now, when we started the task, the mouse didn’t understand a thing.
So they tried to do anything to see if they nailed it.
At first, they moved their whiskers like crazy
to find out if the stick was there, it didn’t care about anything else.
Then they got tired and stopped responding for the rest of the day,
There was no way to motivate them.
And sometimes, when they were motivated,
they’d regain their motivation
and play hard and stick their tongues out all the time
to see if they could win a reward.
(Laughter)
Meanwhile we’d measure the neurons, and we’d see exactly the same:
the neurons in total chaos, very hard to understand what they were doing.
But as we the animal was learning,
the behavior would become increasingly consistent.
They waited for the object until it was positioned,
and just then they moved their whiskers and looked for it very carefully.
Once they detected it, just then, they gave us their answer.
And the neurons also began to behave much more orderly.
The task was divided.
There were neurons that learned to respond
only when they moved their whiskers.
Another group that responded when they made contact with the object.
And another ones that spiked when the mouse stuck out its tongue.
What did we learn from this experiment?
We learned that learning
is a cooperative task at neuronal level.
It requires many neurons to do exactly the same.
Because one day they can disappear and stop responding.
That neurons do exactly the same,
that many do exactly the same,
makes learning robust
and not so dependent on individual neurons.
If a neuron dies, there is another that can come to the rescue.
Also, what we discovered is that,
although learning is very flexible and the neurons are changing
as we learn, this flexibility has limits.
What does this mean?
It means that in the brain there is a built-in structure
that, sometimes, can’t change.
For example, the neuron that responded to the movement of the whiskers
always did exactly the same thing:
no matter the extra learning, it would never happened to be
a neuron that responded to the contact with the object.
Each neuron respected its own task and their own union.
Now, we already know how it works one brain, approximately
when the mouse knows something.
We are closer to considering the greater challenge:
Could we take a new mouse
and artificially modify its neuronal activity in a way
that when we give them a task, they already know how to do it?
That is, can we sculpt neuronal activity
while the mouse is chilled and relaxed,
and that by the first time we give them a task they already knows how to perform?
Well, this is still pretty ambitious,
but we are on the way to achieve it.
In the lab we set out to do something a little simpler:
Can we grab a neuron, a single neuron of the entire brain
and teach it to do something we want?
I’m going to tell you what we did.
We gave the mouse a signal, that time was running
and what they had to do
was to increase the activity of that neuron to a target
and if they did they got a reward.
And how did we do this?
Well, what we did, and here you can see it in the green line,
is to teach the mouse that this neuron had to increase that activity
until it reaches the goal, that peak they see there,
and there they got a reward.
To help the neuron, we gave they a sound.
A lower pitch sound if they were far from the target
and a higher pitch sound if they were closer.
In the same way the mouse did with their whiskers in the beginning,
the mouse didn’t know what to do to change neuronal activity.
By insisting to make the link between the goal —
to increase neuronal activity —
and the reward, the neuron learned to respond faster and faster.
Even though we took a single neuron,
this is the first step so that one day
we can incorporate artificial knowledge in brains.
Also, what we learned from this experiment
is that for the neuron to learn,
the only thing we had to give them feedback information of their state,
tell it how far it was from the goal
and give it a reward when it succeeded.
Basically, to teach this neuron,
we didn’t do anything different to what we do
when we learn something:
we practice a lot, someone tell us if things are right or wrong
and we receive, eventually, a reward if we do it well.
I hope I have convinced you, for now, that the brain
is enormously flexible and that our neural circuits
can change at any time.
Thanks to this, we can learn to use a lot of tools
almost as if they were extensions of our body.
If you know how to play the guitar it’s because in your brain
there are circuits that know what movements the fingers do
to make each note.
For your brain, after all,
your hand is no different than a tool.
Now, if for your brain is not so different
to learn to control your arm than controlling a tool,
does it mean that we could add new limbs?
For example, could we add a module, a chip,
that allows us to control a robotic arm?
Or even crazier, something that adds more learning capacity, more memory,
that comes pre-loaded with languages — Chinese in my case —
or help us do mental calculations much faster?
Although it looks like science fiction, these men and women cyborgs,
are already among us.
They are beginning to do neuronal implants in paraplegic patients
so that patients can control
a robotic arm with their own brain and gain independence.
These implants measure the activity of a few neurons,
decode that information
and control the joints of robotic arms.
Just like before, in the beginning the brain
doesn’t know how to control it, but by effort and practice,
trial and error, it can learn.
What you see here is a patient with one of these implants:
that little box above their head.
With her own brain she is controlling that robotic arm.
Thanks to the huge brain flexibility,
she has managed to change her neural circuits
in a way to control the robotic arm efficiently.
With a lot of practice, movements become more natural
and they begin to be part of her own body.
And, for the first time, in many years,
the patient can help herself a drink.
(Applause)
The final smile of the patient reminds us of how nice
is to learn something new for the first time.
To achieve something considered impossible before.
We live something like this, perhaps, when we are babies
and we get to say our first words and we manage to walk;
as children when we learn to write
and, perhaps, to play soccer;
at twenty-something when we study a career, or a profession.
Sometimes as an adult we don’t have as much patience or time to learn
but if you have the curiosity and you want to learn something new,
remember how nice it is
and don’t doubt that your brain is prepared for that.
Because, after all, if a neuron can learn,
aren’t 80 billion also able to?
Thank you!
(Applause)
”
Be First to Comment