Pawan Sinha on how brains learn to see

譯者: Cherry Huang 審譯者: Tracie Chen

If you are a blind child in India,

如果您是在印度的盲童，

you will very likely have to contend with

您將很有可能必須面臨

at least two big pieces of bad news.

至少兩個重大的壞消息。

The first bad news

第一個壞消息是

is that the chances of getting treatment

得到治療的機會

are extremely slim to none,

幾乎渺茫至零，

and that's because most of the blindness

這是因為在這國家大多數的

alleviation programs in the country

盲人扶助專案

are focused on adults,

都只關注成人，

and there are very, very few hospitals

而只有非常非常少的醫院

that are actually equipped to treat children.

真正有能力治療病童。

In fact, if you were to be treated,

事實上，如果您接受治療，

you might well end up being treated

您可能最後是被

by a person who has no medical credentials

沒有醫療證書的人員治療

as this case from Rajasthan illustrates.

就像拉賈斯坦所揭發的這個實例。

This is a three-year-old orphan girl

這個三歲大孤女

who had cataracts.

患了白內障。

So, her caretakers took her

所以她的看護帶她

to the village medicine man,

到村醫那兒，

and instead of suggesting to the caretakers

然而村醫卻不是建議看護

that the girl be taken to a hospital,

帶小女孩去正規醫院，

the person decided to burn her abdomen

相反地那人決定用燒紅的鐵棒

with red-hot iron bars

炙燙小女孩的腹部

to drive out the demons.

以驅逐惡靈。

The second piece of bad news

接下來第二個壞消息

will be delivered to you

是從神經學家得來，

by neuroscientists, who will tell you

他將告訴您

that if you are older than four or five years of age,

如果您的年紀超過四或五歲

that even if you have your eye corrected,

即使您接受視力矯正

the chances of your brain learning how to see

您大腦學習如何解析所見的機會

are very, very slim --

也十分十分地渺茫

again, slim or none.

再一次地，渺茫或零。

So when I heard these two things,

所以當我知道這兩件事情

it troubled me deeply,

我同時因為個人

both because of personal reasons

以及科學因素

and scientific reasons.

深深感到困擾。

So let me first start with the personal reason.

讓我從個人因素開始說起吧

It'll sound corny, but it's sincere.

這聽起來像陳腔濫調，但是由衷的。

That's my son, Darius.

這是我兒子，Darius。

As a new father,

身為一個新手父親，

I have a qualitatively different sense

對於寶寶有多脆弱

of just how delicate babies are,

我們對他們有多少義務

what our obligations are towards them

還有我們能感覺到對一個孩子

and how much love

有多麼深的愛

we can feel towards a child.

有截然不同而美好的感受

I would move heaven and earth

我可能會為了治療Darius

in order to get treatment for Darius,

在天堂和現實間穿梭。

and for me to be told

對我而言，被告知

that there might be other Dariuses

別的地方還有更多像Darius這樣的孩子

who are not getting treatment,

並沒有得到治療

that's just viscerally wrong.

這直覺地就是個錯誤。

So that's the personal reason.

以上是我的個人因素。

Scientific reason is that this notion

科學方面

from neuroscience of critical periods --

神經科學有個關鍵時期之說

that if the brain is older

也就是，如果大腦年齡大於

than four or five years of age,

四歲或五歲，

it loses its ability to learn --

大腦就喪失學習能力這個觀點。

that doesn't sit well with me,

這無法說服我，

because I don't think that idea

因為我不認為這想法

has been tested adequately.

被充分地驗證過。

The birth of the idea is from

這想法是從

David Hubel and Torsten Wiesel's work,

David Hube和Torsten Wiesel的研究成果孕育而來，

two researchers who were at Harvard,

這兩位科學家曾在哈佛任職

and they got the Nobel Prize in 1981

並且在1981年因為他們視覺生理學

for their studies of visual physiology,

的研究得到諾貝爾獎，

which are remarkably beautiful studies,

那的確是傑出而漂亮的研究，

but I believe some of their work

但我相信其中的一部分成果

has been extrapolated

已被貿然地

into the human domain prematurely.

用於人類領域

So, they did their work with kittens,

他們用不同治療剝奪

with different kinds of deprivation regiments,

的小貓群組進行實驗

and those studies,

而這些可追溯到

which date back to the '60s,

六零年代的研究

are now being applied to human children.

現今被套用於人類孩子上。

So I felt that I needed to do two things.

所以我覺得我需要做兩件事。

One: provide care

一是提供

to children who are currently

目前被剝奪治療的孩子

being deprived of treatment.

醫療照護。

That's the humanitarian mission.

這是基於人道主義的使命。

And the scientific mission would be

而科學的使命則是

to test the limits

試探視覺可塑性

of visual plasticity.

的極限

And these two missions, as you can tell,

你可以看出，這兩個任務

thread together perfectly. One adds to the other;

完美地結合在一起，相輔相成。

in fact, one would be impossible without the other.

事實上是缺一不可。

So, to implement

為了執行

these twin missions,

這個雙生任務，

a few years ago, I launched Project Prakash.

幾年前我開始了Prakash計畫

Prakash, as many of you know,

Prakash，也許你們很多人都知道，

is the Sanskrit word for light,

是梵文中光的總稱

and the idea is that

我們的理念亦是如此

in bringing light into the lives of children,

將光明引入孩子們的生命中，

we also have a chance

同時也有

of shedding light on some of the

一窺某些

deepest mysteries of neuroscience.

神經科學奧秘的機會

And the logo -- even though it looks extremely Irish,

再看看這標誌，雖然它看起來很像幸運草

it's actually derived from

事實上它是源自

the Indian symbol of Diya, an earthen lamp.

一種印度陶燈 Diya 的象徵

The Prakash, the overall effort

Prakash整體工作內容

has three components:

包含三個部分，

outreach, to identify children in need of care;

宣傳理念、定義病童是否需要照護、醫藥治療，

medical treatment; and in subsequent study.

以及後續追蹤研究。

And I want to show you a short video clip

我想要在此分享一段節錄短片

that illustrates the first two components of this work.

來說明前兩個部份的工作。

This is an outreach station

這是一個由為盲童設立的學校

conducted at a school for the blind.

的宣傳站。

(Text: Most of the children are profoundly and permanently blind ...)

(字幕: 大多數的孩童嚴重且永久地失明...

Pawan Sinha: So, because this is a school for the blind,

Pawan Sinha: 因為這是間為盲人而辦的學校，

many children have permanent conditions.

許多盲童是永久地失明。

That's a case of microphthalmos,

這是一個小眼症的例子

which is malformed eyes,

也就是有畸形的眼睛，

and that's a permanent condition;

並且是永久性的。

it cannot be treated.

它沒有辦法被治癒。

That's an extreme of micropthalmos

這和極端的小眼症

called enophthalmos.

稱為眼球陷沒。

But, every so often, we come across children

但偶爾，我們遇見有

who show some residual vision,

殘存視力的孩童

and that is a very good sign

這是很好的徵兆

that the condition might actually be treatable.

顯示病情實際上可能是可治療的。

So, after that screening, we bring the children to the hospital.

於是經過篩檢，我們帶這些病童到醫院去。

That's the hospital we're working with in Delhi,

這是我們在德里工作的醫院，

the Schroff Charity Eye Hospital.

修福慈善眼科醫院

It has a very well-equipped

裏頭有非常裝備完善的

pediatric ophthalmic center,

小兒眼科中心，

which was made possible in part

這得以實現有部分

by a gift from the Ronald McDonald charity.

來自麥當勞叔叔慈善基金的贈與。

So, eating burgers actually helps.

所以吃漢堡的確可助一臂之力。

(Text: Such examinations allow us to improve

(字幕: 這樣的檢查允許我們增進

eye-health in many children, and ...

許多孩童的眼睛健康，並且

... help us find children who can participate in Project Prakash.)

幫助我們找到可參與Prakash計畫的孩子。)

PS: So, as I zoom in to the eyes of this child,

補充一下，當我放大這個孩子眼睛的影像，

you will see the cause of his blindness.

你將可以看到他失明的原因。

The whites that you see in the middle of his pupils

你看到在瞳孔中間白色部分

are congenital cataracts,

是先天性的白內障，

so opacities of the lens.

所以水晶體是不透明的。

In our eyes, the lens is clear,

我們的眼睛水晶體則是透明的。

but in this child, the lens has become opaque,

但是這個孩子的水晶體已經變得不透光，

and therefore he can't see the world.

因此他沒有辦法看見這世界。

So, the child is given treatment. You'll see shots of the eye.

所以這孩子接受治療。你所看到的是眼球置換手術。

Here's the eye with the opaque lens,

這是有不透明水晶體的眼球，

the opaque lens extracted

不透明的水晶體被吸取出

and an acrylic lens inserted.

然後置入丙烯酸人工水晶體。

And here's the same child

這是剛剛那孩子

three weeks post-operation,

在接受手術後三個星期，

with the right eye open.

使用右眼的情形。

(Applause)

(鼓掌)

Thank you.

謝謝大家。

So, even from that little clip, you can begin to get the sense

所以，即使從這小片段，你也能開始感覺到

that recovery is possible,

治癒是可能的，

and we have now

而我們已經

provided treatment to over 200 children,

對超過兩百位孩童施予治療

and the story repeats itself.

而成功復明故事不斷重演。

After treatment, the child

經過治療，孩子

gains significant functionality.

得到重要的視覺官能。

In fact, the story holds true

事實上，這情節即使對你知道的

even if you have a person who got sight

某個被剝奪視力多年的人

after several years of deprivation.

也同樣為真。

We did a paper a few years ago

我們幾年前撰寫了一篇報導

about this woman that you see on the right, SRD,

是有關你看到在右邊，得了SRD(漿液性網膜剝離)的女人，

and she got her sight late in life,

她在晚年復明，

and her vision is remarkable at this age.

而且在這年紀她的視力是相當好的。

I should add a tragic postscript to this --

我在這故事末加個悲劇附錄。

she died two years ago

她兩年前因為一場巴士意外

in a bus accident.

而去世了。

So, hers is just a truly inspiring story --

她的故事正是非常激勵人心的，

unknown, but inspiring story.

不為人知，但激勵人心的故事。

So when we started finding these results,

當我們開始發掘這些成果，

as you might imagine, it created quite a bit of stir

就像你想像的，它在科學以及普及的

in the scientific and the popular press.

出版物造成旋風。

Here's an article in Nature

這是在自然期刊側寫這工作

that profiled this work,

的一篇文章，

and another one in Time.

另一篇則是在時代雜誌。

So, we were fairly convinced -- we are convinced --

我們非常確信，我們確信

that recovery is feasible,

儘管長久地失去視力。

despite extended visual deprivation.

復明是可能的。

The next obvious question to ask:

下一個要問且顯而易見的問題是：

What is the process of recovery?

痊癒的過程如何發生？

So, the way we study that is,

所以，我們研究的方法是，

let's say we find a child who has light sensitivity.

假設我們找到一個對光線有反應的孩子

The child is provided treatment,

提供他治療，

and I want to stress that the treatment

我要強調這治療

is completely unconditional;

是完全無條件的。

there is no quid pro quo.

是無償的。

We treat many more children then we actually work with.

我們治療的孩童比參與計畫的更多。

Every child who needs treatment is treated.

每個需要醫療的孩子都被治療。

After treatment, about every week,

治療後，約每個星期一次，

we run the child

我們讓孩童

on a battery of simple visual tests

接受簡單視覺測試

in order to see how their visual skills

以了解他們看的技能

are coming on line.

是否如預期恢復。

And we try to do this for as long as possible.

我們試著盡可能長時間追蹤。

This arc of development

這些發展曲線，

gives us unprecedented

給了我們有關

and extremely valuable information

如何建構視覺

about how the scaffolding of vision

不同以往且十分有價值

gets set up.

的資訊

What might be the causal connections

早期發展的視覺技能和

between the early developing skills

較晚發展的兩者之間的

and the later developing ones?

因果關聯為何？

And we've used this general approach to study

我們已經用這套公認的研究方法來研究

many different visual proficiencies,

許多不同的視覺熟練程度，

but I want to highlight one particular one,

但我想要強調其中特別的一項，

and that is image parsing into objects.

那就是將圖像解析為物體的技能。

So, any image of the kind that you see on the left,

在這兒，左邊任何一種你所見的圖，

be it a real image or a synthetic image,

是真實影像還是合成的?

it's made up of little regions

這是由中間

that you see in the middle column,

那張有不同顏色、不同照度

regions of different colors, different luminances.

的區域合併起來做成的。

The brain has this complex task

大腦的複雜任務包括

of putting together, integrating,

組合、整合，

subsets of these regions

將這些區域統整

into something that's more meaningful,

成為更有意義的某樣物品，

into what we would consider to be objects,

是我們認知的實物，

as you see on the right.

就像右邊你所看到的那樣。

And nobody knows how this integration happens,

而沒有人知道這些整合工作如何發生。

and that's the question we asked with Project Prakash.

這就是我們在Prakash計畫中所問的問題。

So, here's what happens

這就是發生歷程

very soon after the onset of sight.

一旦開始建構視覺便變化快速。

Here's a person who had gained sight just a couple of weeks ago,

這個人他才在兩個星期前恢復視力，

and you see Ethan Myers, a graduate student from MIT,

再看到Ethan Myers，來自麻省理工的大學生，

running the experiment with him.

正在和他一起做實驗。

His visual-motor coordination is quite poor,

他的視覺運動協調十分地差，

but you get a general sense

但你大約可知

of what are the regions that he's trying to trace out.

他正試著描繪的區域為何

If you show him real world images,

如果向他展示真實世界的影像

if you show others like him real world images,

如果你向其他跟這個人相似的病友展示真實世界的影像，

they are unable to recognize most of the objects

他們不能辨識出大部分的物體

because the world to them is over-fragmented;

因為這樣的世界對他們來說太支離破碎。

it's made up of a collage, a patchwork,

他們所見是由不同顏色、照明的

of regions of different colors and luminances.

小區塊拼貼、補綴而成

And that's what's indicated in the green outlines.

而這就是綠色輪廓所表示的東西。

When you ask them,

當你要求他們，

"Even if you can't name the objects, just point to where the objects are,"

即使他們不能說出物體的名字，僅僅指出在哪裡就好，

these are the regions that they point to.

這些就是他們所指的地方。

So the world is this complex

所以世界是如此複雜的

patchwork of regions.

區塊拼貼。

Even the shadow on the ball

即使是球上面的陰影

becomes its own object.

也是一個獨立的物件。

Interestingly enough,

非常有趣的是，

you give them a few months,

你給他們幾個月的時間，

and this is what happens.

這就是後來所發生的事。

Doctor: How many are these?

醫生：這裡有多少個?

Patient: These are two things.

病人：這裡有兩個東西。

Doctor: What are their shapes?

醫生：他們的形狀是?

Patient: Their shapes ...

病人：他們的形狀...

This one is a circle,

這個是圓的，

and this

這個嘛

is a square.

是方的。

PS: A very dramatic transformation has come about.

附註：非常戲劇化的轉變已經發生。

And the question is:

而問題是：

What underlies this transformation?

這轉變是根據什麼？

It's a profound question,

這是個深奧的問題，

and what's even more amazing is how simple

而令人驚訝的是答案有

the answer is.

多麼地簡單。

The answer lies in motion

答案就是移動

and that's what I want to show you in the next clip.

這也是我想要在下個影片中呈現給大家的。

Doctor: What shape do you see here?

醫生: 你在這兒看到的是什麼形狀?

Patient: I can't make it out.

病患: 我無法回答這個問題。

Doctor: Now?

醫生: 現在呢?

Patient: Triangle.

病患: 三角形。

Doctor: How many things are these?

醫生: 這裡有多少東西?

Now, how many things are these?

現在，這裡有多少東西?

Patient: Two.

病患: 兩個。

Doctor: What are these things?

醫生: 這些東西是什麼呢?

Patient: A square and a circle.

病患: 一個方塊跟一個圓。

PS: And we see this pattern over and over again.

附帶說明，我們看到這個模式不斷重複。

The one thing the visual system needs

視覺系統所需

in order to begin parsing the world

解析世界的要件就是

is dynamic information.

動態資訊。

So the inference we are deriving from this,

從這結果和多次這類的實驗

and several such experiments,

我們推論，

is that dynamic information processing,

動態資訊的處理

or motion processing,

或動作資訊處裡，

serves as the bedrock for building

就像是建立後續複雜的

the rest of the complexity of visual processing;

視覺處理流程的基石。

it leads to visual integration

它成就了視覺整合

and eventually to recognition.

以及最終的辨認能力。

This simple idea has far reaching implications.

這簡單的概念在科學研究上有深遠的含意。

And let me just quickly mention two,

讓我快速地簡介其中兩個。

one, drawing from the domain of engineering,

一是從工程上來說，

and one from the clinic.

另一個是臨床上的意義。

So, from the perspective of engineering,

從工程的透視法，

we can ask: Goven that we know

我們能問，我們已知

that motion is so important for the human visual system,

動作對人類的視覺系統是如此重要，

can we use this as a recipe

我們可以利用這結果建立

for constructing machine-based vision systems

不需要程式設計師

that can learn on their own, that don't need to be programmed

設計程式就可以自我學習的

by a human programmer?

機械視覺系統嗎?

And that's what we're trying to do.

這就是我們正試著達成的。

I'm at MIT, at MIT you need to apply

我在麻省理工，在那兒你需要運用

whatever basic knowledge you gain.

任何你得到的基礎知識。

So we are creating Dylan,

所以我們創造了Dylan

which is a computational system

它是套有遠大目標

with an ambitious goal

的運算系統

of taking in visual inputs

可以接收的和一個人類孩童

of the same kind that a human child would receive,

相同的視覺輸入後，

and autonomously discovering:

自主地發現

What are the objects in this visual input?

這些視覺輸入有什麼東西。

So, don't worry about the internals of Dylan.

不用擔心Dylan裡面有什麼。

Here, I'm just going to talk about

在這裡我只想簡單提到

how we test Dylan.

我們如何測試Dylan。

The way we test Dylan is by giving it

我們測試它的方式是利用

inputs, as I said, of the same kind

給它輸入，像我先前說的，跟一個孩童，

that a baby, or a child in Project Prakash would get.

或是Prakash計畫中的孩子會得到的一樣。

But for a long time we couldn't quite figure out:

但是長久以來我們無法真正知道

Wow can we get these kinds of video inputs?

我們如何能的到這樣的視訊輸入。

So, I thought,

所以，我就想，

could we have Darius

我們是否能使Darius

serve as our babycam carrier,

像我們的嬰兒攝影機攜帶者，

and that way get the inputs that we feed into Dylan?

使我們得以將視訊輸入至Dylan。

So that's what we did.

這就是我們的方法。

(Laughter)

(眾笑)

I had to have long conversations with my wife.

我必須跟我妻子談很久。

(Laughter)

(眾笑)

In fact, Pam, if you're watching this,

事實上，Pam，如果你正看著，

please forgive me.

請原諒我。

So, we modified the optics of the camera

為了更生動模擬寶寶的視力

in order to mimic the baby's visual acuity.

我們修改了攝影機的光學元件。

As some of you might know,

也許你們有些人知道，

babyies are born pretty much legally blind.

寶寶剛出生的時候幾乎是看不見的。

Their acuity -- our acuity is 20/20;

寶寶看20英尺外的東西，

babies' acuity is like 20/800,

就像正常視力在800英尺看見的。

so they are looking at the world

所以他們是非常非常模糊地

in a very, very blurry fashion.

看著這個世界。

Here's what a baby-cam video looks like.

這就是嬰兒攝影機所見的影像。

(Laughter)

(眾大笑)

(Applause)

(鼓掌)

Thankfully, there isn't any audio

謝天謝地，這段影片

to go with this.

沒有任何聲音。

What's amazing is that working with such

令人驚訝的是，即使是這麼

highly degraded input,

高度劣化的影像輸入，

the baby, very quickly, is able

寶寶還是可以很快地

to discover meaning in such input.

知道影像的意義。

But then two or three days afterward,

經過兩三天，

babies begin to pay attention

寶寶開始把注意力放在

to their mother's or their father's face.

他們的爸爸或媽媽臉上。

How does that happen? We want Dylan to be able to do that,

這是如何發生的? 我們希望Dylan可以做到這樣。

and using this mantra of motion,

利用移動的奧秘，

Dylan actually can do that.

事實上Dylan可以做到，

So, given that kind of video input,

即使是那麼糟糕的影像輸入。

with just about six or seven minutes worth of video,

大概只有六或七分鐘可用的影片，

Dylan can begin to extract patterns

Dylan可以開始取得

that include faces.

包含臉部的圖樣。

So, it's an important demonstration

所以這是重要效果

of the power of motion.

的重要展示。

The clinical implication, it comes from the domain of autism.

臨床應用上，要從自閉症開始說起。

Visual integration has been associated with autism

視覺整合已經被好幾位研究學者證實

by several researchers.

和自閉症有關。

When we saw that, we asked:

當我們知道這結果，我們問：

Could the impairment in visual integration

是否動態資訊處理缺損

be the manifestation of something underneath,

就是視覺整合損傷

of dynamic information processing deficiencies in autism?

造成的自閉症的背後肇因

Because, if that hypothesis were to be true,

因為，若是這個假說成立，

it would have massive repercussions in our understanding

這將對我們了解

of what's causing the many different aspects

是什麼使得自閉症有不同方面的外顯特徵

of the autism phenotype.

有重要的回饋。

What you're going to see are

接下來的一段短片是

video clips of two children -- one neurotypical,

兩個孩子在玩乒乓球，一個非自閉症，

one with autism, playing Pong.

另一個有自閉症。

So, while the child is playing Pong, we are tracking where they're looking.

當孩子在玩壁球的時候，我們追蹤他們看著哪兒。

In red are the eye movement traces.

紅色的就是眼球移動的紀錄，

This is the neurotypical child, and what you see

這是非自閉症孩子的，而你所看見的是

is that the child is able to make cues

這個孩子可以利用

of the dynamic information

動態資訊給的暗示

to predict where the ball is going to go.

預測球會往那兒跑。

Even before the ball gets to a place,

即使在球到達之前，

the child is already looking there.

這孩子已經看著那邊。

Contrast this with a child

對照另一個玩著相同遊戲的

with autism playing the same game.

自閉症孩子。

Instead of anticipating,

這個孩子的視線總是

the child always follows where the ball has been.

跟著球到過的地方跑而非預測它。

The efficiency of the use

自閉症患者利用

of dynamic information

動態資訊的效率

seems to be significantly compromised in autism.

似乎明顯地被抑制。

So we are pursuing this line of work

所以我們持續進行這工作

and hopefully we'll have

並且希望我們將很快有

more results to report soon.

更多結果可以報導。

Looking ahead, if you think of this disk

未來，如果你把這圓盤看作

as representing all of the children

我們目前為止

we've treated so far,

治療過的孩子們的數量，

this is the magnitude of the problem.

這就是問題的嚴重程度。

The red dots are the children we have not treated.

這些紅點表示孩子我們尚未能治療的孩子。

So, there are many, many more children who need to be treated,

所以，世界上還有更多的孩子需要治療，

and in order to expand the scope of the project,

為了拓展這個計畫所及的區域，

we are planning on launching

我們正計畫創建

The Prakash Center for Children,

Prakash孩童中心，

which will have a dedicated pediatric hospital,

它將有專門的小兒科醫院，

a school for the children we are treating

一所為治療中的病童設立的學校，

and also a cutting-edge research facility.

以及尖端研究設施。

The Prakash Center will integrate health care,

Prakash中心將真正地

education and research in a way

創造一個真正整合健康照護、

that truly creates the whole

教育以及研究為一體

to be greater than the sum of the parts.

比單純結合各部分更出色的機構。

So, to summarize: Prakash, in its five years of existence,

在此，我做個小結，Prakash，在過去五年裡，

it's had an impact in multiple areas,

已經對許多領域造成衝擊，

ranging from basic neuroscience

範圍從基礎的神經科學裡

plasticity and learning in the brain,

大腦的學習力及可塑性，

to clinically relevant hypotheses like in autism,

到臨床相關的理論，如自閉症，

the development of autonomous machine vision systems,

還有自主性機器視覺系統的發展，

education of the undergraduate and graduate students,

大學生及研究生們的教育，

and most importantly in the alleviation

最重要的，是減少了

of childhood blindness.

盲童的數目。

And for my students and I, it's been

而對我和我的學生來說，這計畫已是

just a phenomenal experience

非比尋常的經驗，

because we have gotten to do interesting research,

因為我們做著有趣的研究，

while at the same time

同時也幫助

helping the many children that we have worked with.

許多參與實驗的孩子們。

Thank you very much.

非常感謝大家

(Applause)

(鼓掌)