in the desert, in the tropical forest

在沙漠中，在熱帶雨林中，

and in my kitchen,

還有我家的廚房，

and in the hills around Silicon Valley where I live.

以及我住的矽谷四周的山丘上。

I've recently realized that ants

我最近才領悟到原來螞蟻

are using interactions differently

使用不同的互動法

in different environments,

在不同的環境中，

and that got me thinking that we could learn from this

而這使我不禁去想， 我們或許能從這件事

about other systems,

了解其他系統，

like brains and data networks that we engineer,

像大腦及我們設計出的數據網路，

and even cancer.

甚至癌症。

So what all these systems have in common

所以這些系統的共通點

is that there's no central control.

在於沒有中央控制。

An ant colony consists of sterile female workers --

蟻群是由不孕的雌性工蟻——

those are the ants you see walking around —

就是你看到四處走動的那些螞蟻——

and then one or more reproductive females

還有一隻或多隻能生育的雌蟻組成，

who just lay the eggs.

而這種雌蟻只下蛋。

They don't give any instructions.

牠們不下任何指令。

Even though they're called queens,

即使牠們被稱為蟻后，

they don't tell anybody what to do.

牠們也不會告訴任何螞蟻要做什麼。

So in an ant colony, there's no one in charge,

所以在蟻群內，沒有負責人。

and all systems like this without central control

所有像這樣沒有中央控制的系統，

are regulated using very simple interactions.

要以非常簡單的互動來規範。

Ants interact using smell.

螞蟻以嗅覺互動。

They smell with their antennae,

牠們以觸角聞味道，

and they interact with their antennae,

而且牠們也以觸角互動，

so when one ant touches another with its antennae,

所以當一隻螞蟻以觸角 碰觸另外一隻螞蟻，

it can tell, for example, if the other ant

牠就能分辨，譬如說這一隻

is a nestmate

是不是同窩蟻，

and what task that other ant has been doing.

還有這隻螞蟻正在做什麼任務。

So here you see a lot of ants moving around

所以這裡你看到很多螞蟻

and interacting in a lab arena

在這個實驗場所四處走動及互動，

that's connected by tubes to two other arenas.

這個場所與另外兩個以管子相連。

So when one ant meets another,

所以當某隻螞蟻碰到另一隻，

it doesn't matter which ant it meets,

碰到的是哪一隻螞蟻並不重要，

and they're actually not transmitting

而且牠們其實並沒有傳送

any kind of complicated signal or message.

任何複雜的信號或訊息。

All that matters to the ant is the rate

對螞蟻而言最重要的是

at which it meets other ants.

牠們碰到其他螞蟻的頻率。

And all of these interactions, taken together,

而所有的互動，全部一起看，

produce a network.

會產生一種網路。

So this is the network of the ants

那麼這就是螞蟻的網路，

that you just saw moving around in the arena,

由你剛剛看到的那個場所 裡面的移動所形成。

and it's this constantly shifting network

就是這個不斷移位的網路，

that produces the behavior of the colony,

產生蟻群的行為，

like whether all the ants are hiding inside the nest,

像是不是所有的螞蟻都躲在窩裡，

or how many are going out to forage.

或有多少在外面覓食。

A brain actually works in the same way,

大腦其實也以同樣的方法運作，

but what's great about ants is

但研究螞蟻最棒的地方，

that you can see the whole network as it happens.

在於你能親眼目睹整個網路形成。

There are more than 12,000 species of ants,

有超過一萬兩千種螞蟻，

in every conceivable environment,

存在於每一種可想到的環境中，

and they're using interactions differently

而牠們互動的方法也不同，

to meet different environmental challenges.

以因應不同的環境挑戰。

So one important environmental challenge

所以有一種很重要的環境挑戰，

that every system has to deal with

是每一種系統都必須面對的，

is operating costs, just what it takes

就是營業成本，就是到底要花多少

to run the system.

來經營系統。

And another environmental challenge is resources,

而另一種環境挑戰則是資源，

finding them and collecting them.

要找尋及收集資源。

In the desert, operating costs are high

在沙漠裡，營業成本很高，

because water is scarce,

因為水很稀少，

and the seed-eating ants that I study in the desert

而且我在沙漠中研究的 一種吃種子的螞蟻，

have to spend water to get water.

必須先用掉水才能得到水。

So an ant outside foraging,

所以一隻螞蟻在外面覓食，

searching for seeds in the hot sun,

在烈日下尋找種子，

just loses water into the air.

就會失去水分，釋放到空氣中。

But the colony gets its water

但這個蟻群會因此得到水份，

by metabolizing the fats out of the seeds

也就是從代謝牠們所吃的種子

that they eat.

所含的脂肪以得到水份，

So in this environment, interactions are used

所以在這樣的環境下，互動是要

to activate foraging.

動員覓食行為。

An outgoing forager doesn't go out unless

指派為覓食者的螞蟻，

it gets enough interactions with returning foragers,

在與回來的覓食蟻 得到足夠的互動前不會出去，

and what you see are the returning foragers

而你現在看到的是回來的覓食蟻，

going into the tunnel, into the nest,

進入隧道中，進入窩裡，

and meeting outgoing foragers on their way out.

與正要出去的覓食蟻互動。

This makes sense for the ant colony,

這種方法對蟻群而言很有道理，

because the more food there is out there,

因為外面的食物愈多，

the more quickly the foragers find it,

覓食蟻就能愈快找到食物，

the faster they come back,

牠們也就愈快回來，

and the more foragers they send out.

所以就會送更多的覓食蟻出去。

The system works to stay stopped,

這個系統的原理是保持不動，

unless something positive happens.

直到有好事發生。

So interactions function to activate foragers.

所以互動的功用 在促使覓食蟻開始活動。

And we've been studying the evolution of this system.

我們一直在研究這個系統的演化。

First of all, there's variation.

首先，這系統裡有變數。

It turns out that colonies are different.

結果證明每一種蟻群都不一樣。

On dry days, some colonies forage less,

在乾燥的日子， 某些蟻群的覓食行為會少一點，

so colonies are different in how

所以每個蟻群的不同點在於

they manage this trade-off

牠們如何權衡

between spending water to search for seeds

是要花掉水分以尋找種子，

and getting water back in the form of seeds.

還是要找種子回來以取得裡面的水。

And we're trying to understand why

我們試圖瞭解為什麼

some colonies forage less than others

某些蟻群的覓食行為較少，

by thinking about ants as neurons,

透過將螞蟻視為神經元的方法，

using models from neuroscience.

並使用神經科學的模式。

So just as a neuron adds up its stimulation

所以就像神經元會累積

from other neurons to decide whether to fire,

從別的神經元送來的刺激， 以決定是否發射，

an ant adds up its stimulation from other ants

螞蟻也會累積 從別的螞蟻傳來的刺激，

to decide whether to forage.

以決定是否出去覓食。

And what we're looking for is whether there might be

我們在找的是，是否在

small differences among colonies

各個蟻群間有微小的差異，

in how many interactions each ant needs

也就是螞蟻需要多少互動

before it's willing to go out and forage,

牠才願意出去覓食，

because a colony like that would forage less.

因為像那樣的蟻群可能較少覓食。

And this raises an analogous question about brains.

這也引發大家對大腦 產生類似的問題。

We talk about the brain,

我們談論大腦，

but of course every brain is slightly different,

但當然每個大腦都有少許不同，

and maybe there are some individuals

而或許有某些人，

or some conditions

或在某些情況下，

in which the electrical properties of neurons are such

神經元的電性質也像這樣，

that they require more stimulus to fire,

需要更多的刺激才能發射，

and that would lead to differences in brain function.

而那會導致大腦功能產生差異。

So in order to ask evolutionary questions,

所以為了要問進化的問題，

we need to know about reproductive success.

我們必須瞭解生殖成功率。

This is a map of the study site

這是研究地點的地圖，

where I have been tracking this population

我已經在那裡追蹤這種

of harvester ant colonies for 28 years,

收割蟻群 28 年了，

which is about as long as a colony lives.

這大約就是一個蟻群的壽命。

Each symbol is a colony,

每一個符號代表一個蟻群，

and the size of the symbol is how many offspring it had,

而符號的大小代表 這個蟻群有多少後代，

because we were able to use genetic variation

因為我們能用遺傳變異

to match up parent and offspring colonies,

去匹配親子蟻群，

that is, to figure out which colonies

也就是說，去找出哪個蟻群

were founded by a daughter queen

是哪一隻蟻后的女兒

produced by which parent colony.

所產生的。

And this was amazing for me, after all these years,

這對我而言很奇妙，在這些年之後，

to find out, for example, that colony 154,

能發現到，舉個例，蟻群 154，

whom I've known well for many years,

多年來我非常瞭解的一個蟻群，

is a great-grandmother.

居然是曾祖母。

Here's her daughter colony,

這是牠的女兒的蟻群，

here's her granddaughter colony,

這是牠的孫女的蟻群，

and these are her great-granddaughter colonies.

而這些是牠的曾孫女的蟻群。

And by doing this, I was able to learn

經由這麼做，我學到

that offspring colonies resemble parent colonies

後代蟻群與親代蟻群，

in their decisions about which days are so hot

在決定哪一天太熱，

that they don't forage,

不出去覓食這方面很相似，

and the offspring of parent colonies

而親代蟻群的後代，

live so far from each other that the ants never meet,

兩者住得很遠， 這些螞蟻從來沒有見過面，

so the ants of the offspring colony

所以後代蟻群的螞蟻，

can't be learning this from the parent colony.

不可能從牠們的親代蟻群學到這個。

And so our next step is to look

因此我們下一步要去找

for the genetic variation underlying this resemblance.

產生這種相似度的遺傳變異。

So then I was able to ask, okay, who's doing better?

所以我才能問，好，誰做得比較好？

Over the time of the study,

在做研究的這段時間，

and especially in the past 10 years,

特別是過去十年，

there's been a very severe and deepening drought

曾有非常嚴重且日益加深的乾旱，

in the Southwestern U.S.,

發生在美國西南部，

and it turns out that the colonies that conserve water,

結果是這些節約用水的蟻群，

that stay in when it's really hot outside,

就是在外面真的 很熱的時候還留在窩裡，

and thus sacrifice getting as much food as possible,

從而犧牲盡可能找最多食物的蟻群，

are the ones more likely to have offspring colonies.

愈有可能產生後代蟻群。

So all this time, I thought that colony 154

所以這些時間，我以為蟻群 154

was a loser, because on really dry days,

是輸家，因為在很乾燥的日子，

there'd be just this trickle of foraging,

牠們只有稀稀落落的覓食行為，

while the other colonies were out

而其他的蟻群都跑出去

foraging, getting lots of food,

覓食，獲得大量的食物，

but in fact, colony 154 is a huge success.

但其實，蟻群 154 非常成功。

She's a matriarch.

她是女族長。

She's one of the rare great-grandmothers on the site.

她是當地少有的曾祖母輩之一。

To my knowledge, this is the first time

據我所知，這也是第一次

that we've been able to track

我們能追蹤

the ongoing evolution of collective behavior

在自然的動物族群中

in a natural population of animals

持續進行的集體行為演化，

and find out what's actually working best.

並找出什麼是最佳的運作方式。

Now, the Internet uses an algorithm

那麼，網際網路使用一種演算法

to regulate the flow of data

以管理資料流，

that's very similar to the one

與這個非常相似，

that the harvester ants are using to regulate

就是收獲蟻在使用以管理

the flow of foragers.

覓食蟻的流程。

And guess what we call this analogy?

你猜我們怎麼叫這種類比？

The anternet is coming.

蟻際網路來了！

(Applause)

（掌聲）

So data doesn't leave the source computer

所以資料不會從源計算機輸出，

unless it gets a signal that there's enough bandwidth

直到它得到信號，有足夠的頻寬

for it to travel on.

讓資料傳出去。

In the early days of the Internet,

在網際網路發展早期，

when operating costs were really high

當營業成本還很高，

and it was really important not to lose any data,

而且很重要 不能失去任何資料的時期，

then the system was set up for interactions

系統被設成互動

to activate the flow of data.

以啟動資料流。

It's interesting that the ants are using an algorithm

有趣的是螞蟻會使用一種演算法，

that's so similar to the one that we recently invented,

與我們最近發明的非常相似，

but this is only one of a handful of ant algorithms

但這只是我們所知的

that we know about,

螞蟻演算法之一罷了，

and ants have had 130 million years

螞蟻有一億三千萬年的時間

to evolve a lot of good ones,

發展很多很好的演算法，

and I think it's very likely

我想非常有可能

that some of the other 12,000 species

在其他一萬二千種螞蟻中，

are going to have interesting algorithms

能找到有趣的演算法

for data networks

給資料網路使用，

that we haven't even thought of yet.

是我們想都沒想過的。

So what happens when operating costs are low?

所以，營業成本低的時候 會發生什麼？

Operating costs are low in the tropics,

熱帶地方的營業成本低，

because it's very humid, and it's easy for the ants

因為那裡很濕，螞蟻很容易

to be outside walking around.

在外面走來走去。

But the ants are so abundant

但在熱帶地方，

and diverse in the tropics

螞蟻量很大，種類極多，

that there's a lot of competition.

因此競爭也很激烈。

Whatever resource one species is using,

某種螞蟻在使用的資源，

another species is likely to be using that

別的螞蟻可能在同時間

at the same time.

也需要使用。

So in this environment, interactions are used

所以在這種環境，互動的使用法

in the opposite way.

恰好相反。

The system keeps going

系統要持續運轉，

unless something negative happens,

直到負面事件發生。

and one species that I study makes circuits

而我在研究的某種螞蟻 還會形成迴路，

in the trees of foraging ants

覓食蟻在蟻窩到食物源

going from the nest to a food source and back,

往來的樹上

just round and round,

不斷繞圈圈，

unless something negative happens,

直到某件負面事件發生為止，

like an interaction

像是與別種螞蟻

with ants of another species.

有了互動。

So here's an example of ant security.

所以這是螞蟻的保全案例。

In the middle, there's an ant

在中間有一隻螞蟻

plugging the nest entrance with its head

正以牠的頭堵住窩的入口，

in response to interactions with another species.

回應牠與另一種螞蟻互動的結果。

Those are the little ones running around

這些小東西跑來跑去

with their abdomens up in the air.

腹部朝上。

But as soon as the threat is passed,

一但威脅消失，

the entrance is open again,

入口就又打開了，

and maybe there are situations

也許在某些情況下，

in computer security

電腦的安全性

where operating costs are low enough

在營運成本夠低時，

that we could just block access temporarily

我們只要暫時把存取擋住，

in response to an immediate threat,

以回應立即性的威脅，

and then open it again,

然後再把它打開就好，

instead of trying to build

而不用試著建立

a permanent firewall or fortress.

一個永久性的防火牆或要塞。

So another environmental challenge

那麼另一種環境挑戰，

that all systems have to deal with

所有系統都要面對的，

is resources, finding and collecting them.

是資源，要找尋及收集資源。

And to do this, ants solve the problem

要做這個，螞蟻解決了

of collective search,

集體搜尋的問題，

and this is a problem that's of great interest

而這在機器人科學

right now in robotics,

是目前極感興趣的問題，

because we've understood that,

因為我們已經瞭解，

rather than sending a single,

與其送出一個單獨操作、

sophisticated, expensive robot out

複雜又很貴的機器人出去

to explore another planet

探險另一個星球，

or to search a burning building,

或去搜索一棟燃燒的建築物，

that instead, it may be more effective

還不如這樣可能更有效，

to get a group of cheaper robots

就是找一組便宜的機器人

exchanging only minimal information,

只交換最少的資訊，

and that's the way that ants do it.

而那正是螞蟻做的方式。

So the invasive Argentine ant

所以入侵種阿根廷蟻

makes expandable search networks.

建造可擴充的搜尋網路。

They're good at dealing with the main problem

牠們對解決集體搜尋的

of collective search,

主要問題很在行，

which is the trade-off between

那是權衡了

searching very thoroughly

要完全徹底的搜尋，

and covering a lot of ground.

還是要涵蓋大片土地的結果。