{♫Intro♫}
{♫Intro♫}

When you think about algorithms, you probably think of Google searches or YouTube recommendations,
When you think about algorithms, you probably think of Google searches or YouTube recommendations,

or predictive text—situations where powerful computers are coming up with information you’re
or predictive text—situations where powerful computers are coming up with information you’re

looking for.
looking for.

An algorithm, though, is basically any recipe of calculations that a computer can follow
An algorithm, though, is basically any recipe of calculations that a computer can follow

to produce a specific kind of information.
to produce a specific kind of information.

And algorithms aren’t just for computers.
And algorithms aren’t just for computers.

They show up all over nature, too, in places like your immune system and in schools of
They show up all over nature, too, in places like your immune system and in schools of

fish.
fish.

And just as engineers borrow ideas from nature’s physical designs, some computer scientists
And just as engineers borrow ideas from nature’s physical designs, some computer scientists

look for inspiration in nature’s algorithms.
look for inspiration in nature’s algorithms.

Here are four ways our technology has improved thanks to algorithms we swiped from nature.
Here are four ways our technology has improved thanks to algorithms we swiped from nature.

Say you’re looking for the perfect fuzzy animal photo to send as a virtual hug to your
Say you’re looking for the perfect fuzzy animal photo to send as a virtual hug to your

friend.
friend.

An image search pulls up some cuddling kittens that are almost right, if only you could find
An image search pulls up some cuddling kittens that are almost right, if only you could find

a slightly more zoomed-out version…
a slightly more zoomed-out version…

What you want in this situation is something called nearest-neighbors search—an algorithm
What you want in this situation is something called nearest-neighbors search—an algorithm

that can quickly search a big database to find the items most similar to one you specify.
that can quickly search a big database to find the items most similar to one you specify.

That gets harder as the database gets bigger; and on the internet, there are way too many
That gets harder as the database gets bigger; and on the internet, there are way too many

images for the search engine to compare your photo with every single one.
images for the search engine to compare your photo with every single one.

So how do search engines pull off that feature that gives you “visually similar images”?
So how do search engines pull off that feature that gives you “visually similar images”?

One technique is called locality-sensitive hashing.
One technique is called locality-sensitive hashing.

This is a type of algorithm that digests each image into a short digital fingerprint called
This is a type of algorithm that digests each image into a short digital fingerprint called

a hash, with similar hashes for similar inputs.
a hash, with similar hashes for similar inputs.

For example, if your inputs were essays, a decent hash might be the first letters of
For example, if your inputs were essays, a decent hash might be the first letters of

the first twenty sentences.
the first twenty sentences.

So if one essay was copied from another, their hashes would likely be very close.
So if one essay was copied from another, their hashes would likely be very close.

This method makes it easy to find similar inputs.
This method makes it easy to find similar inputs.

Instead of comparing your kittens to every other image on the internet, Google can organize
Instead of comparing your kittens to every other image on the internet, Google can organize

images by their hashes and just pull out the similar ones.
images by their hashes and just pull out the similar ones.

The catch is that locality-sensitive hashing can still be kind of slow, and sometimes inaccurate.
The catch is that locality-sensitive hashing can still be kind of slow, and sometimes inaccurate.

That is where fly brains come to the rescue.
That is where fly brains come to the rescue.

See, a fly can smell, but it doesn’t differentiate every subtle variation of odor; it groups
See, a fly can smell, but it doesn’t differentiate every subtle variation of odor; it groups

odors into categories so it can learn that, for instance, cheese smells often lead to
odors into categories so it can learn that, for instance, cheese smells often lead to

fruit, but book smells don’t.
fruit, but book smells don’t.

In 2017, a team of computer scientists and biologists realized that fly brains group
In 2017, a team of computer scientists and biologists realized that fly brains group

odors using a form of locality-sensitive hashing.
odors using a form of locality-sensitive hashing.

Except in the flies’ version, the brain boils a smell down to a few numbers by first
Except in the flies’ version, the brain boils a smell down to a few numbers by first

expanding the smell data into a much larger collection of numbers.
expanding the smell data into a much larger collection of numbers.

Only then does it select a few of those numbers as the hash.
Only then does it select a few of those numbers as the hash.

It’s sort of like expanding an essay by replacing each character with a random 10-character
It’s sort of like expanding an essay by replacing each character with a random 10-character

code, producing a string of gibberish ten times as long.
code, producing a string of gibberish ten times as long.

Then you could find the hundred gibberish words that appear most frequently, take the
Then you could find the hundred gibberish words that appear most frequently, take the

first letter, and use that as the essay’s hash.
first letter, and use that as the essay’s hash.

As strange as that strategy sounds, it turns out to work really well.
As strange as that strategy sounds, it turns out to work really well.

All the extra gobbledegook gives the algorithm more opportunities to find patterns that jump
All the extra gobbledegook gives the algorithm more opportunities to find patterns that jump

out strongly for one cluster of inputs but are conspicuously absent for others.
out strongly for one cluster of inputs but are conspicuously absent for others.

When the computer scientists built their own fly-based hashing algorithm, it was up to
When the computer scientists built their own fly-based hashing algorithm, it was up to

twice as accurate as traditional methods—and also twenty times faster!
twice as accurate as traditional methods—and also twenty times faster!

Computer vision is everywhere.
Computer vision is everywhere.

Self-driving cars, MRI technology, facial recognition; they all use it.
Self-driving cars, MRI technology, facial recognition; they all use it.

Most of these systems need to do some form of object recognition—meaning they need
Most of these systems need to do some form of object recognition—meaning they need

to identify the contents of an image.
to identify the contents of an image.

For decades, computer scientists used handcrafted algorithms to extract image features like
For decades, computer scientists used handcrafted algorithms to extract image features like

edges and contiguous shapes.
edges and contiguous shapes.

Then, they could then build other algorithms that used those features to guess what was
Then, they could then build other algorithms that used those features to guess what was

in each part of an image.
in each part of an image.

But all these hand-tuned algorithms tend to be fragile.
But all these hand-tuned algorithms tend to be fragile.

It’s up to the cleverness of engineers to design the right kinds of analysis and tweak
It’s up to the cleverness of engineers to design the right kinds of analysis and tweak

the parameters just so.
the parameters just so.

Now, engineers are pretty clever, but there’s only so much subtlety and detail they can
Now, engineers are pretty clever, but there’s only so much subtlety and detail they can

code up.
code up.

In the background, though, a different approach was taking shape: convolutional neural networks,
In the background, though, a different approach was taking shape: convolutional neural networks,

or CNNs.
or CNNs.

In artificial intelligence, most kinds of neural networks are based on nature only in
In artificial intelligence, most kinds of neural networks are based on nature only in

a crude way.
a crude way.

Like, they’re called neural networks because they kind of work like neurons.
Like, they’re called neural networks because they kind of work like neurons.

But convolutional neural networks are based on Nobel Prize-winning research on cat brains.
But convolutional neural networks are based on Nobel Prize-winning research on cat brains.

Back in the 1950s, a pair of neuroscientists discovered that some neurons in a cat’s
Back in the 1950s, a pair of neuroscientists discovered that some neurons in a cat’s

visual cortex, called simple cells, would respond only to simple visual elements—like
visual cortex, called simple cells, would respond only to simple visual elements—like

a line in a specific place at a specific orientation.
a line in a specific place at a specific orientation.

Those simple cells pass information to so-called complex cells, which aggregate the information
Those simple cells pass information to so-called complex cells, which aggregate the information

across a wider area.
across a wider area.

In other words, these researchers discovered a hierarchy in the brain’s visual processing:
In other words, these researchers discovered a hierarchy in the brain’s visual processing:

Earlier layers detect basic features at different locations, then later layers add all that
Earlier layers detect basic features at different locations, then later layers add all that

together to detect more complex patterns.
together to detect more complex patterns.

That structure directly inspired the first convolutional neural networks.
That structure directly inspired the first convolutional neural networks.

In the first layer of a CNN, each simulated neuron looks only at one small patch of the
In the first layer of a CNN, each simulated neuron looks only at one small patch of the

image and checks how well that matches a simple template, like a spot of blue or an edge between
image and checks how well that matches a simple template, like a spot of blue or an edge between

light and dark.
light and dark.

The neuron gives the patch a score depending on how well that patch matches the neuron’s
The neuron gives the patch a score depending on how well that patch matches the neuron’s

template.
template.

Then, the next level looks at all the scores for edges and spots in a slightly bigger patch
Then, the next level looks at all the scores for edges and spots in a slightly bigger patch

and matches them against a more complex template, and so on up the hierarchy until you’re
and matches them against a more complex template, and so on up the hierarchy until you’re

looking for cat paws and bicycle wheels.
looking for cat paws and bicycle wheels.

A CNN learns these templates automatically from data, saving engineers from manually
A CNN learns these templates automatically from data, saving engineers from manually

specifying what to look for.
specifying what to look for.

Today, CNNs totally dominate computer vision.
Today, CNNs totally dominate computer vision.

And although they now have bells and whistles that have nothing to do with the brain, the
And although they now have bells and whistles that have nothing to do with the brain, the

visual hierarchy is still baked in.
visual hierarchy is still baked in.

Next, companies really hate getting hacked.
Next, companies really hate getting hacked.

There are lawsuits and bad press, and it’s pretty inconvenient for them and the people
There are lawsuits and bad press, and it’s pretty inconvenient for them and the people

who rely on them.
who rely on them.

So if a company’s network starts getting hammered with unusual traffic, it might be
So if a company’s network starts getting hammered with unusual traffic, it might be

a good idea to lock things down.
a good idea to lock things down.

But detecting what counts as unusual traffic isn’t always easy.
But detecting what counts as unusual traffic isn’t always easy.

It’s an example of what’s called anomaly detection, or scanning for atypical data,
It’s an example of what’s called anomaly detection, or scanning for atypical data,

which can be tricky.
which can be tricky.

See, you can’t just lay out rules for what normal traffic looks like.
See, you can’t just lay out rules for what normal traffic looks like.

For one thing, what is normal is always changing.
For one thing, what is normal is always changing.

And anyway, hard rules would be too rigid: You wouldn’t want a red alert before every
And anyway, hard rules would be too rigid: You wouldn’t want a red alert before every

holiday just because a bunch of employees traveled early and logged in from home.
holiday just because a bunch of employees traveled early and logged in from home.

It might be tempting to try supervised machine learning, where you show an algorithm lots
It might be tempting to try supervised machine learning, where you show an algorithm lots

of good and bad examples, and it figures out how to tell them apart.
of good and bad examples, and it figures out how to tell them apart.

But with anomaly detection, you often don’t have many examples of the bad stuff you’re
But with anomaly detection, you often don’t have many examples of the bad stuff you’re

trying to catch!
trying to catch!

Most of what a company has, of course, is logs of normal network traffic.
Most of what a company has, of course, is logs of normal network traffic.

So how can it learn what abnormal traffic looks like?
So how can it learn what abnormal traffic looks like?

One particularly cool solution is based on our bodies.
One particularly cool solution is based on our bodies.

Because you know what’s really good at detecting a few bad guys in a sea of things that belong?
Because you know what’s really good at detecting a few bad guys in a sea of things that belong?

Your immune system.
Your immune system.

To recognize and kill off invaders, your immune system uses cells called lymphocytes, which
To recognize and kill off invaders, your immune system uses cells called lymphocytes, which

have little receptors that detect foreign proteins.
have little receptors that detect foreign proteins.

But your body actually produces a huge variety of lymphocytes, with receptors that detect
But your body actually produces a huge variety of lymphocytes, with receptors that detect

pretty much any random protein snippet—including bits of proteins that are supposed to be around.
pretty much any random protein snippet—including bits of proteins that are supposed to be around.

You don’t want to attack those, so before your body releases its lymphocytes, your thymus
You don’t want to attack those, so before your body releases its lymphocytes, your thymus

gland selectively kills off the ones that would detect familiar proteins.
gland selectively kills off the ones that would detect familiar proteins.

As a result, the only lymphocytes that survive are ones that detect foreign proteins.
As a result, the only lymphocytes that survive are ones that detect foreign proteins.

This is called negative selection, and anomaly-detection algorithms can use a similar concept to spot
This is called negative selection, and anomaly-detection algorithms can use a similar concept to spot

unusual traffic.
unusual traffic.

They can generate detectors for random sequences of traffic data, then delete any detectors
They can generate detectors for random sequences of traffic data, then delete any detectors

that go off on normal traffic logs.
that go off on normal traffic logs.

The ones that remain, thus, respond only to abnormal patterns.
The ones that remain, thus, respond only to abnormal patterns.

Finally, in lots of situations, having multiple computers coordinate to divide up a task is
Finally, in lots of situations, having multiple computers coordinate to divide up a task is

crucial—for example, to carry out a robotic search-and-rescue mission, or to index the
crucial—for example, to carry out a robotic search-and-rescue mission, or to index the

entire internet.
entire internet.

When you have just a few computers in a network, it’s easy to have one central command computer
When you have just a few computers in a network, it’s easy to have one central command computer

coordinate them all.
coordinate them all.

But if you’re coordinating hundreds of thousands of machines, or the machines are cut off from
But if you’re coordinating hundreds of thousands of machines, or the machines are cut off from

one another, controlling them with one central computer becomes impractical.
one another, controlling them with one central computer becomes impractical.

So all those machines need a process that they all follow independently that somehow
So all those machines need a process that they all follow independently that somehow

gets the job done efficiently and without horrible mistakes.
gets the job done efficiently and without horrible mistakes.

Little machines… acting independently… getting big projects done… sounds… kind
Little machines… acting independently… getting big projects done… sounds… kind

of like an insect colony!
of like an insect colony!

As it happens, there’s a whole niche of what are called swarm intelligence algorithms
As it happens, there’s a whole niche of what are called swarm intelligence algorithms

that tackle problems like this, and many are based on insect behavior.
that tackle problems like this, and many are based on insect behavior.

For example, there are construction robots that collaborate by imitating termites.
For example, there are construction robots that collaborate by imitating termites.

We still don’t know exactly how termites build their massive mounds.
We still don’t know exactly how termites build their massive mounds.

But we do know that each worker can only see its local environment—what’s been built
But we do know that each worker can only see its local environment—what’s been built

right there and where surrounding workers are.
right there and where surrounding workers are.

That means the only way for the termites to coordinate is by leaving indirect signals
That means the only way for the termites to coordinate is by leaving indirect signals

for each other in their shared environment.
for each other in their shared environment.

Like, when one termite does a bit of construction work, it leaves the soil arranged as some
Like, when one termite does a bit of construction work, it leaves the soil arranged as some

kind of cue to other termites about what needs to be done next.
kind of cue to other termites about what needs to be done next.

This indirect coordination strategy is called stigmergy.
This indirect coordination strategy is called stigmergy.

Inspired by termite stigmergy, a system of robots called TERMES allows a fleet of little
Inspired by termite stigmergy, a system of robots called TERMES allows a fleet of little

robots to build arbitrary structures with no central coordination.
robots to build arbitrary structures with no central coordination.

Just by sensing what’s been built and following some basic traffic rules, each robot figures
Just by sensing what’s been built and following some basic traffic rules, each robot figures

out what to do next to get closer to the target structure.
out what to do next to get closer to the target structure.

The hope is for similar robots to one day build complex structures even in hostile environments,
The hope is for similar robots to one day build complex structures even in hostile environments,

like war zones or on Mars, without depending on a centralized controller.
like war zones or on Mars, without depending on a centralized controller.

Now, nature-inspired algorithms can get a bit out of hand.
Now, nature-inspired algorithms can get a bit out of hand.

People have designed algorithms based on wolf pack behavior, virus evolution, lightning
People have designed algorithms based on wolf pack behavior, virus evolution, lightning

paths, and on and on.
paths, and on and on.

Nature-inspired computing has been criticized for encouraging cute metaphors that don’t
Nature-inspired computing has been criticized for encouraging cute metaphors that don’t

add insight or are unnecessarily complicated.
add insight or are unnecessarily complicated.

But as you’ve seen, sometimes natural phenomena really can make for great inspiration.
But as you’ve seen, sometimes natural phenomena really can make for great inspiration.

Nature can be quite the computer scientist, if you just know where to look!
Nature can be quite the computer scientist, if you just know where to look!

Thanks for watching this episode of SciShow!
Thanks for watching this episode of SciShow!

And if you want to learn more about the amazing ways that nature is inspiring modern technology,
And if you want to learn more about the amazing ways that nature is inspiring modern technology,

you might like our episode on how animals are helping us build better drones.
you might like our episode on how animals are helping us build better drones.

You can watch that right after this.
You can watch that right after this.

{♫Outro♫}
{♫Outro♫}