How the Genes and the Brain See the World | Daeyeol Lee | TEDxKFAS


Translator: Ellen Maloney
Reviewer: Enid Kim Hi, I’m a neuroscientist. I study how the brain makes decisions. Everyday decisions, like what to eat, how much to eat, when to sleep, how much work you want to do, how much money you want to save, etc. etc. And as in many other
scientific areas of research scientists have made
a lot of amazing discoveries in the last couple of decades
about this problem. We have a pretty good idea
as to how the brain makes decisions. But as you can imagine,
this is awfully complicated. But don’t worry; I’m not going to talk about that today. Instead, I’d like to tell you
about something deeper. Something more fundamental
and a bit more metaphysical questions. How well do we make decisions? How good are we as decision makers? Because this is the key to smart living. And if you think about it, we are unhappy sometimes
because we make bad decisions. And more often than not, we are unhappy because we think other
people make bad decisions. Okay? So to understand better why we are not perfect
in making decisions, I think we actually have to look
at the relationship between genes and the brain. As previous speakers have mentioned, genes are the segments of DNA molecules
that includes the instructions to build our entire body,
including our brain. And therefore, we can get some insight
as to the failures of decision-making that takes place in the brain
if we understand this relationship better. And to get some insight as to the relationship
between the genes and the brain, I’d like to tell you about this machine A robot, called “Curiosity”, that engineers of NASA
built a couple of years ago, and sent to Mars. NASA engineers and scientists
had one mission; they wanted to understand
the environment of Mars. But they cannot go there. It’s too far away. So they built the robot
and sent it there instead. But there is a problem. Because Mars is so far away,
these engineers cannot control these machine directly
using a joystick, for example. Imagine a scenario
where this robot, Curiosity, is driving down the field
and it discovers a cliff. It cannot wait for the “stop” command
to arrive from the earth. Because if it sends an image to the earth, it takes about 12 minutes,
on average, to get to the earth. And then if the scientists at NASA
issued a command to stop the robot, it will take about the same amount of time
to come back to the robot. So it would be too late. The robot would be destroyed
if it was waiting for the command. So it needs an ability
to make decisions autonomously, i.e. it needs artificial intelligence. In fact, NASA engineers realized this, and they wrote more than
2.5 lines of computer codes to create an artificial intelligence, and installed it in this robot
before it was sent to Mars. So what happens here
is that engineers have delegated the decision-making
capabilities to the robot. And this happens
all the time in our lives. We frequently delegate a task,
important tasks, to other people when we ask someone else
to do something for you. This is called delegation. However, there are two major
problems with delegation. First kind of problem
is called “moral hazard”. And this happens because
the person who has the original task, is referred to as a principal, gives a request to the agent
who has to follow that instruction. But the agent can cheat. I think everyone in this room
was a child at some point. And we all know
how difficult it is for parents to control their children’s behavior. And we also all remember
how easy it was for a child to cheat; do something else other than
the original instruction they received from the parents. This is a very common problem
of moral hazard. And that’s because action is uncertain. What the agent is going to do
is uncertain to the principal. And it also happens in many companies. CEOs usually want
their employees to work hard, but the employees can cheat. There’s a moral hazard. Insurance companies want their consumers
to be as careful as possible, but once the consumers buy the insurance,
they can be less careful. And that is a lot of monetary loss
for the insurance company. So that’s one problem; moral hazard. The other problem
is called “adverse selection”, and this is the opposite problem. It happens because sometimes the agent follows the instructions
from the principal too carefully. And that’s, in some cases, not desirable because the agent
who receives the instruction or requests from the principal, almost always has more information, more accurate information
about the environment. So sometimes a child, your child,
may know better about their environment. And in those cases, if they follow your instructions
too carefully or blindly, they can actually do harm. And this is true for employee/employer
relationship as well. Sometimes employees at the company
might come up with really creative ideas, and you have to harvest them, rather than just giving them
word-by-word instructions. And I want you to imagine: what if the same problems exist between the genes and the brain? Because after all, brains are built based upon
the instructions encoded in the genes. Okay? So now I’m going to tell you
what kind of problems the brain faces in order to make good decisions
and why it has a difficulty. And I’m going to tell you about
three different problems as an example. So the first problem is that the brain doesn’t have an idea
as to what’s the best. Of course, if you don’t know what’s best,
you can’t make good decisions. You cannot make rational decisions because rationality,
at least by economists, is defined as doing the best. But the brain doesn’t know
what’s the best. So how does the brain deal with this? Finding what’s the best,
among engineers and scientists, is known as optimization. And when they have a difficult problem
of optimization, they have shortcuts. They do something called gradient descent
and it’s a very simple procedure. What you do is that you pick two options and then compare which one’s better
and then pick the one that’s better. And you repeat this
as many times as possible. And since gradually you’re improving
an option that you’re considering, you’ll get closer and closer
to the best solution over time. That’s what’s called gradient descent
and this is what the brain does. This is how the brain tries to find
the best option possible. And in order to do that, the brain uses three
different kinds of baselines. In other words, rather than
constantly picking two random options. It has a baseline, and then
it picks different options, and then wait until you find something
that’s better than your current baseline. So what kind of baseline
does the brain use? One very common baseline
that we all use all the time is your past experience. And this is a good baseline. Because the fact
that you’re all here today means that you successfully avoided
making really bad mistakes in your past. Because if you did, you wouldn’t be here. (Laughter) You would have been dead. So the fact that you’re all alive, breathing, listening to my talk,
means that your past was pretty good. Not only that, your ancestors’ past
was also really, really good. So it sounds like a pretty good idea, using your past
as a baseline for comparison in order to find the best solutions. Like a really good idea. But there is a problem. This doesn’t work all the time. There are cases where
this is not a good strategy. When? When everything’s getting worse. The world is constantly fluctuating and therefore, if you use
your past as a baseline and try to find something that’s better
when everything’s getting worse, that becomes an impossible task. Therefore you get constantly
frustrated for no good reason. Because the brain’s goal
is not necessarily to make things better for the future
but to find something that’s best. Okay? So this is a good heuristic
but sometimes you have to give it up. The second, and I should tell you that the brain
has implemented this strategy by using a chemical called dopamine, the chemical shown on the left, and this is a chemical that increases whenever the brain
finds something better than the past, and this chemical decreases whenever you find something
worse than the past. And this signal is broadcast
to the entire brain, which tells you that the brain
takes this strategy very seriously. But this is not the only baseline
that the brain uses. Another baseline that you often use
is called “hindsight”. You take an action, see what happens from
the action that you select, but after that, in many cases,
you also have the opportunity to observe what would’ve happened
had you done something different. And you can use that as a baseline because if the action
that you did not take seemed to have produced better action, this is what’s called hindsight. Then when you face
the same situation in the future, you can change your mind, you can do something else. And this is what we’ve produced
by emotional response called regret. We have regrets a lot, because sometimes
you realize after the fact that we could have done something better. But just like the first baseline,
this strategy is not perfect. Because there are many cases
where we cannot go back to the past and make the same decision again. Often, we face unique choices
and you have to go beyond that. And in those cases, worrying too much
about what you could have done, how you could have done
better is also wasteful, and leads to unnecessary frustration. So that emotional energy is wasted. And in fact, there seems to be
a dedicated system in the brain called orbital frontal cortex, O.F.C., that plays a special role
in this comparison because somehow, if you had
an unfortunate situation and lose that piece of the brain, then you may lose the ability
to regret entirely, which may be convenient in some cases. There is a third kind of baseline
that the brain uses often. And that is to compare what you did
with what other people are getting. And this is also very powerful because a lot of us face
the same kind of situations together. And to realize that somebody else
is doing better than you should be a really useful information because it tells you what you
could have done to do better. So just like the other two baselines, this is a very useful
baseline in most cases. But if you use it too much
it also could be bad. Because sometimes you cannot do
what somebody else did. After all, I’m not Einstein. So I cannot make fundamental
physical discoveries. Most of you are not as rich as Bill Gates. And, therefore, your options
are much more limited than what Bill Gates can do. And in those cases, comparing you with someone else is wasteful, and leads
to unnecessary frustrations. So it’s not a good strategy
in those cases. So the fact that the brain uses a baseline
to make lots of comparisons in order to find what’s the best approach that the brains can come up with
from using the instructions from the gene. But you have to realize
that they are not perfect. Second kind of problem
that the brain faces, which is also really challenging, are social problems. In general, decision making
in a social context is much harder because you have to make predictions
as to how other people will behave. And that’s a very complicated problem. So how does the brain deal with this? It tries to solve this problem by assuming that everything in your life
has some human-like qualities. We tend to treat everything
that we encounter as if it’s a part of me. And this has many different consequences. And this is why, for example, when we look at a face of a panda, we have an illusion that we can understand
the mind of a panda, because we assume
that the panda is like us. And it saves you time
because we don’t have to decide whether the panda is some
completely different animal, whether it has human-like qualities. It works really well when
we’re interacting with other people, which is what we, most of the time, do. But this sometimes can go too far. For example, when you see a robot
that looks like a human, we tend to begin to treat
those robots like humans, sometimes, forgetting about the problems
that our fellow humans have. And that’s not ideal. There are other problems
that the brain has. And the examples that I gave you so far already implies that the solutions that the brain
comes up with to deal with all the decision-making
problems that we face are multiple. We don’t have one, unified solution. We have many different solutions
implemented sometimes in different parts of the brain. And this creates the last problem
that I wanted to tell you about. And that is the problem of arbitration. To give you an example
of what is arbitration, I want you to imagine
the following hypothetical problem. Imagine that you have an upcoming vacation and you have an opportunity
to visit either London or Paris. Where would you go? How would you decide? You can try to solve
this problem by studying various facts about these two cities, by reading books about London and Paris, or by reading Wikipages
about London and Paris, and find out visiting which city
would be more pleasant. But there are other strategies, too. You can bypass all of that
and just go to a local travel agency, and ask the travel agent
which city would be more interesting. But what if there are
multiple travel agencies? Then you have to choose. You have to decide which travel
agency you should go to, and this is a meta-selection problem. And we face these kinds
of problems all the time as well. And unlike the first cases where you have to compare
the benefits and costs of different cities now, you have to compare the reputations
of different travel agencies. And that’s a completely different problem. Fortunately for us, human brain evolution
has created a system in the front of our brain that specializes in solving
this arbitration problem. And I think that’s our hope
because you don’t have to worry. There is a system in your brain that evaluates the past successes
of different strategies, and that system is designed to guide you
to make a good meta-selection, so that you can use appropriate
decision-making strategies when you encounter a different problem. So to sum up, I have two main messages for you. One is that the brain
is not the principal. The brain is the agent. It does not create
the instructions on its own. It gets the instructions from the gene, which means that it can face
adverse selection problem. In other words, you may discover
facts and information that the brain, the genes,
the principal did not know about. And it’s up to you to use
that information efficiently to make smart decisions. The second one is that we have
multiple strategies in learning algorithms in our brain. We’re not programmed
to work with one simple strategy. Our mind is not unified
as you might think. It has many different solutions. And the world is changing constantly. Therefore, you should be flexible, and you should not be stuck
with one strategy. Thank you for your attention. (Applause)

4 Comments

Add a Comment

Your email address will not be published. Required fields are marked *