A brain in a supercomputer | Henry Markram

翻译人员: dahong zhang 校对人员: Yuanxuan Wang

Our mission is to build

我们的任务是建立

a detailed, realistic

一个复杂的，可行的

computer model of the human brain.

人脑计算机模型。

And we've done, in the past four years,

过去四年，我们已经

a proof of concept

在一小块鼠脑上

on a small part of the rodent brain,

进行概念验证，

and with this proof of concept we are now scaling the project up

根据这样的概念验证，我们正扩大该项目规模到

to reach the human brain.

人脑大小。

Why are we doing this?

我们为什么要这样做呢？

There are three important reasons.

有三个重要原因。

The first is, it's essential for us to understand the human brain

首先，了解人类的大脑对我们来说非常重要

if we do want to get along in society,

如果我们要继续在社会中生存

and I think that it is a key step in evolution.

同时我认为也是进化上的关键一步。

The second reason is,

第二个原因是，

we cannot keep doing animal experimentation forever,

我们不能总是进行动物试验，

and we have to embody all our data and all our knowledge

我们要把我们所有的数据和知识包含进

into a working model.

一个可用模型中。

It's like a Noah's Ark. It's like an archive.

好比诺亚方舟，好比一个档案馆。

And the third reason is that there are two billion people on the planet

第三个原因是：地球上有20亿人

that are affected by mental disorder,

患有精神疾病，

and the drugs that are used today

他们的用药

are largely empirical.

主要依靠经验。

I think that we can come up with very concrete solutions on

我认为，我们能就如何对待疾病的

how to treat disorders.

有非常具体的解决办法。

Now, even at this stage,

现在，即使在这个阶段，

we can use the brain model

我们可以利用大脑模型

to explore some fundamental questions

探讨一些基本的问题

about how the brain works.

关于大脑是如何工作的。

And here, at TED, for the first time,

这里，首次通过TED

I'd like to share with you how we're addressing

我想和大家分享我们如何解决

one theory -- there are many theories --

一个理论 -- 有许多理论 --

one theory of how the brain works.

一个关于大脑如何工作的理论。

So, this theory is that the brain

这种理论认为大脑

creates, builds, a version of the universe,

创建，构建了一个版本的宇宙。

and projects this version of the universe,

把这个宇宙作映射，

like a bubble, all around us.

向气泡一样，映射到我们周围。

Now, this is of course a topic of philosophical debate for centuries.

当然，这是数百年哲学辩论的题目。

But, for the first time, we can actually address this,

但是，有史以来，我们实际上可以解决这个问题，

with brain simulation,

依靠大脑仿真，

and ask very systematic and rigorous questions,

同时提出非常系统和严谨的问题，

whether this theory could possibly be true.

这个理论是否有可能是正确的。

The reason why the moon is huge on the horizon

之所以月亮在地平线上是巨大的

is simply because our perceptual bubble

纯粹是因为我们的知觉气泡

does not stretch out 380,000 kilometers.

没有延伸到三十八万公里外。

It runs out of space.

用光了空间。

And so what we do is we compare the buildings

所以我们比较建筑物

within our perceptual bubble,

在我们的知觉气泡范围内，

and we make a decision.

我们做出一个判断。

We make a decision it's that big,

我们做出判断，那个建筑很大，

even though it's not that big.

即使它并没有那么大，

And what that illustrates

这表明

is that decisions are the key things

判断是支持

that support our perceptual bubble. It keeps it alive.

知觉气泡的关键。判断维系了知觉气泡。

Without decisions you cannot see, you cannot think,

没有判断你不能看，不能思考，

you cannot feel.

不能感知。

And you may think that anesthetics work

没准你会认为麻醉药的工作原理是

by sending you into some deep sleep,

让你熟睡，

or by blocking your receptors so that you don't feel pain,

或者阻断感知器官，使你感不到疼痛，

but in fact most anesthetics don't work that way.

不过，实际上多数麻醉药不是这么工作的。

What they do is they introduce a noise

麻醉药引入干扰

into the brain so that the neurons cannot understand each other.

到大脑中，这样神经元互相之间不能理解。

They are confused,

神经元糊涂了，

and you cannot make a decision.

你也就不能做出一个判断。

So, while you're trying to make up your mind

所以，当你正试图弥补你的思绪的，

what the doctor, the surgeon, is doing

以为手术医生

while he's hacking away at your body, he's long gone.

正在做手术，实际上医生早走了。

He's at home having tea.

正在家里喝茶呢。

(Laughter)

（笑）

So, when you walk up to a door and you open it,

当你拾阶而上打开一扇门的时候，

what you compulsively have to do to perceive

你不由自主的

is to make decisions,

做出判断，

thousands of decisions about the size of the room,

成千上万的判断，诸如房间的大小，

the walls, the height, the objects in this room.

墙壁，高度，房内物体。

99 percent of what you see

99%的你所见，

is not what comes in through the eyes.

并非来自你的眼睛。

It is what you infer about that room.

是你对这个房间推断。

So I can say, with some certainty,

所以我可以确定的说，

"I think, therefore I am."

“我思故我在。“

But I cannot say, "You think, therefore you are,"

但我不能说，“你思故你在。”

because "you" are within my perceptual bubble.

因为你包含在我的知觉气泡中。

Now, we can speculate and philosophize this,

我们可以就此进行哲学思考，

but we don't actually have to for the next hundred years.

但我们没必要再思考上百年。

We can ask a very concrete question.

我们可以提出非常具体的问题。

"Can the brain build such a perception?"

“大脑能建立这样一个知觉吗？”

Is it capable of doing it?

有能力做吗?

Does it have the substance to do it?

大脑有物质创造这些吗?

And that's what I'm going to describe to you today.

这就是我今天要讲的东西。

So, it took the universe 11 billion years to build the brain.

宇宙用了110亿年时间创造的大脑。

It had to improve it a little bit.

宇宙一点点的完善大脑。

It had to add to the frontal part, so that you would have instincts,

增加了额叶，所以有了本能，

because they had to cope on land.

因为要适应陆地生活。

But the real big step was the neocortex.

不过真正的进步是大脑皮层。

It's a new brain. You needed it.

一个新大脑。你需要它。

The mammals needed it

哺乳动物需要它

because they had to cope with parenthood,

因为哺乳动物需要适应父母的角色，

social interactions,

社会互动，

complex cognitive functions.

复杂的认知功能。

So, you can think of the neocortex

你可以认为大脑皮层

actually as the ultimate solution today,

是现在最终解决方案，

of the universe as we know it.

就我们所知的宇宙。

It's the pinnacle, it's the final product

大脑皮层是顶峰，是宇宙的

that the universe has produced.

最终产品。

It was so successful in evolution

在进化上很成功

that from mouse to man it expanded

从老鼠到人类

about a thousandfold in terms of the numbers of neurons,

扩展千倍的神经元细胞，

to produce this almost frightening

来制造这吓人的

organ, structure.

器官，组织。

And it has not stopped its evolutionary path.

它的进化没有停止。

In fact, the neocortex in the human brain

实际上，人脑中的大脑皮层

is evolving at an enormous speed.

还在飞快的进化中。

If you zoom into the surface of the neocortex,

如果你放大到大脑皮层表面，

you discover that it's made up of little modules,

你会发现它有许多小模块组成，

G5 processors, like in a computer.

就像计算机里的G5处理器。

But there are about a million of them.

只不过，大脑皮层有上百万块。

They were so successful in evolution

这些模块在进化上非常成功

that what we did was to duplicate them

我们所作的

over and over and add more and more of them to the brain

就是在大脑中不断的复制这些模块

until we ran out of space in the skull.

直到我们用光了头骨中的空间。

And the brain started to fold in on itself,

大脑开始折叠自身，

and that's why the neocortex is so highly convoluted.

这就是为什么大脑皮层如此褶皱。

We're just packing in columns,

我们只有成列的包装，

so that we'd have more neocortical columns

这样才能有更多的皮层

to perform more complex functions.

实现更复杂的功能。

So you can think of the neocortex actually as

你可以把大脑皮层想想为

a massive grand piano,

一台巨大的钢琴，

a million-key grand piano.

由上百万个琴键的钢琴。

Each of these neocortical columns

每一个皮层

would produce a note.

发出一个音符。

You stimulate it; it produces a symphony.

你刺激它；大脑就奏出交响乐。

But it's not just a symphony of perception.

不只是知觉的交响乐。

It's a symphony of your universe, your reality.

是你的宇宙，你的现实的交响乐。

Now, of course it takes years to learn how

现在，当然需要很多年

to master a grand piano with a million keys.

才能掌握有上百万个琴键的钢琴。

That's why you have to send your kids to good schools,

这就是为什么你送你的孩子上好学校，

hopefully eventually to Oxford.

希望最终上牛津。

But it's not only education.

但不只是教育。

It's also genetics.

这也是遗传。

You may be born lucky,

可能你生来幸运，

where you know how to master your neocortical column,

或者你知道如何掌握皮层。

and you can play a fantastic symphony.

你可以奏出美妙的交响乐。

In fact, there is a new theory of autism

实际上，有一个新的自闭症理论

called the "intense world" theory,

叫做“紧张世界”理论，

which suggests that the neocortical columns are super-columns.

提出皮层列是超级列。

They are highly reactive, and they are super-plastic,

它们超有活性，可塑性，

and so the autists are probably capable of

这样自闭症者可能有能力

building and learning a symphony

建立并学习一首交响乐

which is unthinkable for us.

而对普通人来说不可想象。

But you can also understand

你也可以理解

that if you have a disease

如果你有一种疾病

within one of these columns,

在这些皮层列中，

the note is going to be off.

音符关闭了。

The perception, the symphony that you create

知觉，也就是你创造的交响乐

is going to be corrupted,

也随之跑调，

and you will have symptoms of disease.

你会有疾病的症状。

So, the Holy Grail for neuroscience

所以，神经科学的圣杯

is really to understand the design of the neocoritical column --

就是理解皮层的设计 --

and it's not just for neuroscience;

不只对神经科学；

it's perhaps to understand perception, to understand reality,

可能对理解知觉，理解现实有帮助，

and perhaps to even also understand physical reality.

还可能对理解物理现实又帮助。

So, what we did was, for the past 15 years,

所以过去15年里，

was to dissect out the neocortex, systematically.

我们所作的就是系统的解刨大脑皮层。

It's a bit like going and cataloging a piece of the rainforest.

就像是去分类一片雨林。

How many trees does it have?

有多少树？

What shapes are the trees?

树的形状？

How many of each type of tree do you have? Where are they positioned?

每种树有多少棵？在哪里？

But it's a bit more than cataloging because you actually have to

不过可能比分类还要复杂，因为需要你

describe and discover all the rules of communication,

描述并发现所有的通讯规律，

the rules of connectivity,

连接的规律，

because the neurons don't just like to connect with any neuron.

因为神经元不是随便和别的神经元连接的。

They choose very carefully who they connect with.

神经元挑选连接对象非常谨慎。

It's also more than cataloging

比分类还复杂的是

because you actually have to build three-dimensional

因为需要你建立三维

digital models of them.

数字模型。

And we did that for tens of thousands of neurons,

我们为数万个神经元

built digital models of all the different types

建立了数字模型，我所见到的

of neurons we came across.

不同种类的神经元。

And once you have that, you can actually

一旦你有了这些，你就可以

begin to build the neocortical column.

建立皮层列。

And here we're coiling them up.

这里，我们把它们缠绕在一起。

But as you do this, what you see

这样做的时候，你所见的

is that the branches intersect

就是分叉的相交

actually in millions of locations,

实践上再数百万的点上。

and at each of these intersections

每个交叉

they can form a synapse.

都形成一个突触

And a synapse is a chemical location

每个突触就是一个化学位置

where they communicate with each other.

这里可以进行相互通讯。

And these synapses together

这些突触一起

form the network

构成了网络

or the circuit of the brain.

或者说是大脑的电路

Now, the circuit, you could also think of as

电路，也可以想象为

the fabric of the brain.

大脑的编制结构。

And when you think of the fabric of the brain,

当你想到大脑的编制结构时，

the structure, how is it built? What is the pattern of the carpet?

结构，如何构建？这个编织物的图案是什么？

You realize that this poses

你意识到这个构成

a fundamental challenge to any theory of the brain,

是对任何大脑理论的挑战，

and especially to a theory that says

特别对一种理论，说

that there is some reality that emerges

现实浮现

out of this carpet, out of this particular carpet

出这个编织物，

with a particular pattern.

以一种特有的图案。

The reason is because the most important design secret of the brain

原因是大脑重要的设计秘密是

is diversity.

多样性。

Every neuron is different.

每个神经元都不同。

It's the same in the forest. Every pine tree is different.

如同在森林中。每棵松树都不同。

You may have many different types of trees,

可以有许多同类的书，

but every pine tree is different. And in the brain it's the same.

但是每棵松树都不一样。大脑也一样。

So there is no neuron in my brain that is the same as another,

我大脑中的每个神经元都不相同，

and there is no neuron in my brain that is the same as in yours.

我大脑中的神经元也不会和你的一样。

And your neurons are not going to be oriented and positioned

你的大脑中的神经元

in exactly the same way.

以一样与众不同。

And you may have more or less neurons.

你可能有或多或少的神经元。

So it's very unlikely

所以不可能

that you got the same fabric, the same circuitry.

有同样的编制结构，同样的电路。

So, how could we possibly create a reality

这里，我们如何建立现实

that we can even understand each other?

而又彼此可以理解的呢？

Well, we don't have to speculate.

我们没有必要推测。

We can look at all 10 million synapses now.

现在我们可以看看这一千万突触。

We can look at the fabric. And we can change neurons.

我们可以看看编制结构。我们可以变更神经元。

We can use different neurons with different variations.

我可以用不同的神经元。

We can position them in different places,

我们可以把怕们摆放到不同的位置上，

orient them in different places.

朝向不同的方位。

We can use less or more of them.

我们可以用或多或少的神经元。

And when we do that

我们这么做了之后，

what we discovered is that the circuitry does change.

我们发现，电路却是变化了。

But the pattern of how the circuitry is designed does not.

但是电路设计的模式没有变化。

So, the fabric of the brain,

所以，大脑的编制结构，

even though your brain may be smaller, bigger,

不管你的大脑或大或小，

it may have different types of neurons,

神经元的种类不同，

different morphologies of neurons,

神经元的形态不同，

we actually do share

我们实际上共享了

the same fabric.

同样的编制结构。

And we think this is species-specific,

我们认为这是物种特异性的，

which means that that could explain

这意味着可以解释

why we can't communicate across species.

为什么我们不能和别的物种交流。

So, let's switch it on. But to do it, what you have to do

好，打开它。你必须作的是

is you have to make this come alive.

使它活起来。

We make it come alive

我们让它们活过来

with equations, a lot of mathematics.

用数学公式，很多的数学运算。

And, in fact, the equations that make neurons into electrical generators

实际上，这些让神经元变成发电机的数学公式

were discovered by two Cambridge Nobel Laureates.

是被两个剑桥的诺贝尔奖得主发现的。

So, we have the mathematics to make neurons come alive.

我们又让神经元活过来的数学。

We also have the mathematics to describe

我们也有

how neurons collect information,

神经元采集信息的数学描述，

and how they create a little lightning bolt

以及它们如何制造闪电

to communicate with each other.

来和彼此通讯的。

And when they get to the synapse,

当到达突触后，

what they do is they effectively,

它们有效的，

literally, shock the synapse.

正确的，震动突触。

It's like electrical shock

像是电击

that releases the chemicals from these synapses.

从这些突触中释放化学物质。

And we've got the mathematics to describe this process.

我们有数学工具描述这一过程。

So we can describe the communication between the neurons.

我们可以描述神经元之间的彼此通讯。

There literally are only a handful

真正的少数

of equations that you need to simulate

数学公式用来仿真

the activity of the neocortex.

大脑皮层的活动。

But what you do need is a very big computer.

不过你得要一个非常大的计算机。

And in fact you need one laptop

实际上一台笔记本电脑

to do all the calculations just for one neuron.

只够计算一个神经元用的。

So you need 10,000 laptops.

所以得要一万台笔记本电脑。

So where do you go? You go to IBM,

所以需要IBM帮助，

and you get a supercomputer, because they know how to take

那里有超级计算机，

10,000 laptops and put it into the size of a refrigerator.

可以把一万台笔记本电脑压缩到一个冰箱大笑。

So now we have this Blue Gene supercomputer.

所以我们有蓝色基因超级计算机。

We can load up all the neurons,

我们可以加载所有的神经元，

each one on to its processor,

每个在一个处理器上，

and fire it up, and see what happens.

然后运行，看看发生了什么。

Take the magic carpet for a ride.

骑上魔毯。

Here we activate it. And this gives the first glimpse

我们激活它。看一看

of what is happening in your brain

当有刺激的时候

when there is a stimulation.

我们的大脑在做什么。

It's the first view.

这是第一次看到。

Now, when you look at that the first time, you think,

当你第一次看到的时候，你会想，

"My god. How is reality coming out of that?"

“老天。现实就从这里面产生的？“

But, in fact, you can start,

实际上，你可以开始

even though we haven't trained this neocortical column

即使我们没有训练这个皮层列

to create a specific reality.

来创建一个特定现实。

But we can ask, "Where is the rose?"

不过我们可以问，“玫瑰在哪里？”

We can ask, "Where is it inside,

我们可以问，"在那里，

if we stimulate it with a picture?"

如果我们用一幅图来刺激它？“

Where is it inside the neocortex?

在大脑皮层哪里？

Ultimately it's got to be there if we stimulated it with it.

如果我们刺激它，最终它会在那里。

So, the way that we can look at that

我们看待的方法

is to ignore the neurons, ignore the synapses,

就是忽略神经元，忽略突触，

and look just at the raw electrical activity.

只看原始的电活动。

Because that is what it's creating.

因为这就是它创建的。

It's creating electrical patterns.

创建电模式。

So when we did this,

当我们这么做时，

we indeed, for the first time,

我们实际上第一次看到，

saw these ghost-like structures:

看到这些幽灵一样的结构：

electrical objects appearing

电物体出现

within the neocortical column.

在皮层列中。

And it's these electrical objects

这些电物体

that are holding all the information about

控制了所有刺激它的

whatever stimulated it.

信息。

And then when we zoomed into this,

当我们放大的时候，

it's like a veritable universe.

确实像个宇宙。

So the next step

下面一步是

is just to take these brain coordinates

提出这些大脑中的坐标

and to project them into perceptual space.

投射的知觉空间中。

And if you do that,

如果你这样做，

you will be able to step inside

你就可以走进

the reality that is created

所创造的现实中

by this machine,

由这个机器创造的，

by this piece of the brain.

通过这一块大脑。

So, in summary,

总的来说，

I think that the universe may have --

我认为宇宙可以

it's possible --

可能

evolved a brain to see itself,

演化出能看到自己的大脑，

which may be a first step in becoming aware of itself.

这是意识到自身的第一步。

There is a lot more to do to test these theories,

还有很多需要做的来测试这个理论，

and to test any other theories.

以及其他理论。

But I hope that you are at least partly convinced

不过我希望你可以至少部分的相信

that it is not impossible to build a brain.

建造大脑不是不可能的。

We can do it within 10 years,

我们可以在10年内完成，

and if we do succeed,

如果我们成功了，

we will send to TED, in 10 years,

10年内我们会发到TED，

a hologram to talk to you. Thank you.

一个全息图。谢谢。

(Applause)

（掌声）