English-Video.net comment policy

The comment field is common to all languages

Let's write in your language and use "Google Translate" together

Please refer to informative community guidelines on TED.com

TED2015

Laura Schulz: The surprisingly logical minds of babies

羅拉·舒爾茨: 嬰兒令人驚訝的邏輯思維

Filmed
Views 1,632,838

嬰兒如何從零開始,快速地學到這麼多的事情? 在這個有趣和引用了很多實驗的講座中,認知科學家羅拉·舒爾茨分享嬰兒如何在懂得說話前,已經能用令人驚訝的邏輯思維作出決定。

- Cognitive scientist
Developmental behavior studies spearheaded by Laura Schulz are changing our notions of how children learn. Full bio

馬克.吐溫用一句妙語概括了,
我認為是認知科學的一個最根本問題。
00:12
Mark標記 Twain吐溫 summed總結 up
what I take to be
00:14
one of the fundamental基本的 problems問題
of cognitive認知 science科學
00:18
with a single witticism妙語.
他說:「科學的有趣之處在於,
00:20
He said, "There's something
fascinating迷人 about science科學.
一個人可從微不足道的事得出了偉大的猜想。」
00:23
One gets得到 such這樣 wholesale批發
returns回報 of conjecture推測
00:26
out of such這樣 a trifling輕微
investment投資 in fact事實."
(笑聲)
00:29
(Laughter笑聲)
馬克.吐溫當然只是開玩笑,但他是對的。
00:32
Twain吐溫 meant意味著 it as a joke玩笑,
of course課程, but he's right:
科學有其有趣之處。
00:34
There's something
fascinating迷人 about science科學.
從幾塊骨頭,我們推測了恐龍的存在;
00:37
From a few少數 bones骨頭, we infer推斷
the existence存在 of dinosuarsdinosuars.
從譜線得出了星雲的成份;
00:42
From spectral光譜 lines,
the composition組成 of nebulae星雲.
從果蠅得出了遺傳的機制;
00:47
From fruit水果 flies蒼蠅,
00:50
the mechanisms機制 of heredity遺傳,
以及從血液流入大腦的重建影像,
00:53
and from reconstructed重建 images圖片
of blood血液 flowing流動 through通過 the brain,
在我的研究則是從幼兒的行為中,
00:57
or in my case案件, from the behavior行為
of very young年輕 children孩子,
我們嘗試解釋人類認知的基本機制。
01:02
we try to say something about
the fundamental基本的 mechanisms機制
01:05
of human人的 cognition認識.
我在麻省理工大腦及認知科學系實驗室中,
01:07
In particular特定, in my lab實驗室 in the Department
of Brain and Cognitive認知 Sciences科學 at MITMIT,
花了過去十年研究一個謎團,
01:12
I have spent花費 the past過去 decade
trying to understand理解 the mystery神秘
就是兒童如何從零開始,
快速地學到那麼多的東西。
01:16
of how children孩子 learn學習 so much
from so little so quickly很快.
科學令人著迷之處,
01:20
Because, it turns out that
the fascinating迷人 thing about science科學
亦正是孩子令人著迷的地方。
01:23
is also a fascinating迷人
thing about children孩子,
回應馬克.吐溫的話,
01:27
which哪一個, to put a gentler溫和
spin on Mark標記 Twain吐溫,
那就是孩子從零碎和離亂的訊息中,
能夠得出豐富而抽象的推論的能力。
01:29
is precisely恰恰 their ability能力
to draw rich豐富, abstract抽象 inferences推論
01:34
rapidly急速 and accurately準確
from sparse, noisy嘈雜 data數據.
我將舉出兩個例子:
01:40
I'm going to give you
just two examples例子 today今天.
一個是關於廣義化的問題,
01:42
One is about a problem問題 of generalization概括,
另一個則是關於因果推理的。
01:45
and the other is about a problem問題
of causal因果 reasoning推理.
雖然我將會談及我實驗室的研究,
01:47
And although雖然 I'm going to talk
about work in my lab實驗室,
但這個研究的靈感是來自一個領域,
01:50
this work is inspired啟發 by
and indebted感激的 to a field領域.
一個我要感謝世界各地的導師、
同事和工作夥伴付出的領域。
01:53
I'm grateful感激 to mentors導師, colleagues同事,
and collaborators合作者 around the world世界.
讓我先談談廣義化的問題。
01:59
Let me start開始 with the problem問題
of generalization概括.
歸納數據樣本在科學上是不可或缺的,
02:02
Generalizing泛化 from small samples樣本 of data數據
is the bread麵包 and butter牛油 of science科學.
如我們調查一部分的選民,
02:06
We poll輪詢 a tiny fraction分數 of the electorate選民
然後預測國家大選的結果。
02:09
and we predict預測 the outcome結果
of national國民 elections選舉.
我們觀察一小撮病人在臨床試驗中的反應,
02:12
We see how a handful少數 of patients耐心
responds響應 to treatment治療 in a clinical臨床 trial審訊,
然後把藥物帶入市場,
02:16
and we bring帶來 drugs毒品 to a national國民 market市場.
但只有在整個人口中隨機抽樣才可行。
02:19
But this only works作品 if our sample樣品
is randomly隨機 drawn from the population人口.
當我們刻意挑選樣本,
02:23
If our sample樣品 is cherry-picked櫻桃採摘
in some way --
如我們只調查城市中的選民,
02:26
say, we poll輪詢 only urban城市的 voters選民,
又或在治療心臟病的臨床試驗中,
02:28
or say, in our clinical臨床 trials試驗
for treatments治療 for heart disease疾病,
我們只研究男性,
02:32
we include包括 only men男人 --
這樣的結果便不能代表整個人口。
02:34
the results結果 may可能 not generalize概括
to the broader更廣泛 population人口.
因此科學家著緊抽樣的方法是否隨機。
02:38
So scientists科學家們 care關心 whether是否 evidence證據
is randomly隨機 sampled取樣 or not,
但這又跟嬰兒有甚麼關係?
02:42
but what does that have to do with babies嬰兒?
嬰兒在任何時候都要歸納數據樣本,
02:44
Well, babies嬰兒 have to generalize概括
from small samples樣本 of data數據 all the time.
當他們看到幾隻橡皮鴨,
並知道它們浮在水面。
02:49
They see a few少數 rubber橡膠 ducks鴨子
and learn學習 that they float浮動,
又或見到幾個皮球,
並知道它們能彈跳。
02:52
or a few少數 balls and learn學習 that they bounce彈跳.
從中他們建立對橡膠鴨和皮球的概念,
02:55
And they develop發展 expectations期望
about ducks鴨子 and balls
並將這概念延伸至日後會見到的
所有橡膠鴨和皮球。
02:58
that they're going to extend延伸
to rubber橡膠 ducks鴨子 and balls
03:01
for the rest休息 of their lives生活.
嬰兒對橡膠鴨和皮球的這種概括,
03:03
And the kinds of generalizations概括
babies嬰兒 have to make about ducks鴨子 and balls
他們會運用在每一件事上:
03:07
they have to make about almost幾乎 everything:
鞋子、船、封蠟、捲心菜和皇帝。
03:09
shoes and ships船舶 and sealing密封 wax
and cabbages捲心菜 and kings國王.
因此嬰兒留意這些細節能否代表整體。
03:14
So do babies嬰兒 care關心 whether是否
the tiny bit of evidence證據 they see
03:17
is plausibly振振有詞 representative代表
of a larger population人口?
我們一起看看吧。
03:21
Let's find out.
我將讓你看兩段短片,
03:23
I'm going to show顯示 you two movies電影,
這兩段短片分別代表實驗的兩個情況。
03:25
one from each of two conditions條件
of an experiment實驗,
由於你將看到兩段短片,
03:27
and because you're going to see
just two movies電影,
你只會看到兩個嬰兒,
03:30
you're going to see just two babies嬰兒,
而這兩個嬰兒在很多地方都是不同的。
03:32
and any two babies嬰兒 differ不同 from each other
in innumerable無數 ways方法.
但這兩個嬰兒將代表更大的群組,
03:36
But these babies嬰兒, of course課程,
here stand in for groups of babies嬰兒,
你將看到的不同之處則代表
嬰兒行為中的平均差異。
03:39
and the differences分歧 you're going to see
03:41
represent代表 average平均 group differences分歧
in babies'嬰兒 behavior行為 across橫過 conditions條件.
在每一段短片中你會見到嬰兒
在做些他們正常會做的事。
03:47
In each movie電影, you're going to see
a baby寶寶 doing maybe
03:49
just exactly究竟 what you might威力
expect期望 a baby寶寶 to do,
嬰兒本身已是十分神奇的,
03:53
and we can hardly幾乎不 make babies嬰兒
more magical神奇 than they already已經 are.
但對我來說他們的神奇之處,
03:58
But to my mind心神 the magical神奇 thing,
也是我想你們留意的地方,
04:00
and what I want you to pay工資 attention注意 to,
就是這兩種情況之間的分別。
04:02
is the contrast對比 between之間
these two conditions條件,
因為這兩段短片唯一不同的地方,
04:05
because the only thing
that differs不同 between之間 these two movies電影
正是嬰兒將要觀察的資料。
04:08
is the statistical統計 evidence證據
the babies嬰兒 are going to observe.
我們把一些藍色和黃色的球給嬰兒看。
04:13
We're going to show顯示 babies嬰兒
a box of blue藍色 and yellow黃色 balls,
權孝媛當時是我的學生,
現在則是史丹佛大學的同事。
04:16
and my then-graduate那麼研究生 student學生,
now colleague同事 at Stanford斯坦福, HyowonHyowon GweonGweon,
她將拿出三個藍色的球,
04:21
is going to pull three blue藍色 balls
in a row out of this box,
而當她拿出這些球時,
她會把球擠一下,
04:24
and when she pulls those balls out,
she's going to squeeze them,
讓這些球發出吱吱聲。
04:27
and the balls are going to squeak.
這對於嬰兒來說就像TED一樣,
04:29
And if you're a baby寶寶,
that's like a TEDTED Talk.
是件很美好的事。
04:32
It doesn't get better than that.
(笑聲)
04:34
(Laughter笑聲)
從一個裝滿藍色球的箱中,
抽出三個藍色球是件很容易的事。
04:38
But the important重要 point is it's really
easy簡單 to pull three blue藍色 balls in a row
04:42
out of a box of mostly大多 blue藍色 balls.
你閉上眼睛也能做到,
04:44
You could do that with your eyes眼睛 closed關閉.
這就像隨機抽樣。
04:46
It's plausibly振振有詞 a random隨機 sample樣品
from this population人口.
因此當你可以在箱中隨機地抽出
能吱吱叫的物件時,
04:49
And if you can reach達到 into a box at random隨機
and pull out things that squeak,
也許箱中所有物件都能吱吱叫,
04:53
then maybe everything in the box squeaks尖叫聲.
所以嬰兒可能會假設黃色球也能吱吱叫。
04:56
So maybe babies嬰兒 should expect期望
those yellow黃色 balls to squeak as well.
但這些黃色球都有一根棒,
05:00
Now, those yellow黃色 balls
have funny滑稽 sticks on the end結束,
所以嬰兒可用它們做些不同的事,
05:02
so babies嬰兒 could do other things
with them if they wanted to.
他們可以拍打或搖動這些球。
05:05
They could pound them or whack重打 them.
就讓我們看看這嬰兒會做甚麼。
05:07
But let's see what the baby寶寶 does.
(影片) 權孝媛: 看看這個。
(球發出吱吱聲)
05:12
(Video視頻) HyowonHyowon GweonGweon: See this?
(Ball squeaks尖叫聲)
看到這個嗎?
(球發出吱吱聲)
05:16
Did you see that?
(Ball squeaks尖叫聲)
很酷吧!
05:20
Cool.
看看這個。
05:24
See this one?
(球發出吱吱聲)
05:26
(Ball squeaks尖叫聲)
哇!
05:28
Wow.
羅拉·舒爾茨: 早就說了。 (笑聲)
05:33
Laura勞拉 Schulz舒爾茨: Told you. (Laughs)
(影片) 孝媛: 看到這個嗎?
(球發出吱吱聲)
05:35
(Video視頻) HGHG: See this one?
(Ball squeaks尖叫聲)
克拉拉, 這個是給你的, 你拿去玩吧。
05:39
Hey Clara克拉拉, this one's那些 for you.
You can go ahead and play.
(笑聲)
05:51
(Laughter笑聲)
羅拉: 我不用解釋, 對吧?
05:56
LSLS: I don't even have to talk, right?
嬰兒把藍色球的特性套用到黃色球上。
05:59
All right, it's nice不錯 that babies嬰兒
will generalize概括 properties性能
06:02
of blue藍色 balls to yellow黃色 balls,
嬰兒從模仿我們中學習,這是很神奇的,
06:03
and it's impressive有聲有色 that babies嬰兒
can learn學習 from imitating冒充 us,
但我們早就知道嬰兒能這樣做。
06:06
but we've我們已經 known已知 those things about babies嬰兒
for a very long time.
有趣的地方是當把一樣的東西給嬰兒看時,
甚麼事會發生。
06:10
The really interesting有趣 question
06:12
is what happens發生 when we show顯示 babies嬰兒
exactly究竟 the same相同 thing,
我們能肯定這是完全一樣的,
因我們有個秘密的空間,
06:15
and we can ensure確保 it's exactly究竟 the same相同
because we have a secret秘密 compartment隔室
從中我們抽出這些球。
06:18
and we actually其實 pull the balls from there,
但這次我們改變了抽樣的母體。
06:20
but this time, all we change更改
is the apparent明顯的 population人口
06:24
from which哪一個 that evidence證據 was drawn.
這次我們在一個裝滿黃色球的箱中,
抽出三個藍色球給嬰兒看。
06:27
This time, we're going to show顯示 babies嬰兒
three blue藍色 balls
06:30
pulled out of a box
of mostly大多 yellow黃色 balls,
想想甚麼事會發生?
06:34
and guess猜測 what?
你大概不能隨機地在裝滿黃色球的箱中,
連續抽出三個藍色球,
06:35
You [probably大概 won't慣於] randomly隨機 draw
three blue藍色 balls in a row
06:38
out of a box of mostly大多 yellow黃色 balls.
06:40
That is not plausibly振振有詞
randomly隨機 sampled取樣 evidence證據.
因此這很可能不是隨機抽樣。
這反映了孝媛可能是刻意抽出藍色球,
06:44
That evidence證據 suggests提示 that maybe HyowonHyowon
was deliberately故意 sampling採樣 the blue藍色 balls.
可能這些藍色球是特別的,
06:49
Maybe there's something special特別
about the blue藍色 balls.
可能只有藍色球能吱吱叫。
06:52
Maybe only the blue藍色 balls squeak.
一起看看這嬰兒會做甚麼。
06:55
Let's see what the baby寶寶 does.
(影片) 孝媛: 看看這個。
(球發出吱吱聲)
06:57
(Video視頻) HGHG: See this?
(Ball squeaks尖叫聲)
看到這個玩具嗎? (球發出吱吱聲)
07:02
See this toy玩具?
(Ball squeaks尖叫聲)
哇, 這很酷, 看到嗎? (球發出吱吱聲)
07:05
Oh, that was cool. See?
(Ball squeaks尖叫聲)
這個是給你的, 你拿去玩吧。
07:10
Now this one's那些 for you to play.
You can go ahead and play.
(不耐煩的)
(笑聲)
07:18
(Fussing大驚小怪)
(Laughter笑聲)
羅拉: 你剛剛看到兩個15月大的嬰兒,
07:26
LSLS: So you just saw
two 15-month-old個月大 babies嬰兒
按他們觀察到樣本出現的機率,
而做出完全不同的事。
07:29
do entirely完全 different不同 things
07:31
based基於 only on the probability可能性
of the sample樣品 they observed觀察到的.
07:35
Let me show顯示 you the experimental試驗 results結果.
一起看看實驗的結果,
垂直軸代表在每一個情況中,
有多少百分比的嬰兒擠壓球。
07:37
On the vertical垂直 axis, you'll你會 see
the percentage百分比 of babies嬰兒
07:40
who squeezed擠壓 the ball in each condition條件,
你可看見嬰兒在樣本和整體一致時,
07:42
and as you'll你會 see, babies嬰兒 are much
more likely容易 to generalize概括 the evidence證據
比刻意挑選的樣本,
07:46
when it's plausibly振振有詞 representative代表
of the population人口
較會歸納他們看到的特徵。
07:49
than when the evidence證據
is clearly明確地 cherry-picked櫻桃採摘.
因此這帶出一個有趣的預測。
07:53
And this leads引線 to a fun開玩笑 prediction預測:
假設你在一個裝滿黃色球的箱中
只拿出一個藍色球,
07:55
Suppose假設 you pulled just one blue藍色 ball
out of the mostly大多 yellow黃色 box.
當然你很難隨機地連續抽出三個藍色球,
08:00
You [probably大概 won't慣於] pull three blue藍色 balls
in a row at random隨機 out of a yellow黃色 box,
但你可以只用一個藍色球作樣本,
08:04
but you could randomly隨機 sample樣品
just one blue藍色 ball.
這不一定是個不可行的樣本。
08:07
That's not an improbable難以置信 sample樣品.
當你隨機抽出一個會吱吱叫的東西時,
08:09
And if you could reach達到 into
a box at random隨機
可能箱中所有的東西都會吱吱叫,
08:11
and pull out something that squeaks尖叫聲,
maybe everything in the box squeaks尖叫聲.
因此雖然嬰兒會看到較少吱吱叫的例子,
08:15
So even though雖然 babies嬰兒 are going to see
much less evidence證據 for squeaking吱吱,
而且在只抽出一個球的情況下,
他們會有較少的動作去模仿,
08:20
and have many許多 fewer actions行動 to imitate模擬
08:22
in this one ball condition條件 than in
the condition條件 you just saw,
08:25
we predicted預料到的 that babies嬰兒 themselves他們自己
would squeeze more,
但我們預計會有更多嬰兒擠壓球。
這正是我們發現的結果。
08:29
and that's exactly究竟 what we found發現.
因此15月大的嬰兒在這方面就像科學家,
08:32
So 15-month-old個月大 babies嬰兒,
in this respect尊重, like scientists科學家們,
他們留意抽樣的方法是否隨機,
08:37
care關心 whether是否 evidence證據
is randomly隨機 sampled取樣 or not,
並以此建立對事物的概念:
08:40
and they use this to develop發展
expectations期望 about the world世界:
甚麼會吱吱叫而甚麼不會,
08:43
what squeaks尖叫聲 and what doesn't,
甚麼需要探索而甚麼可忽略。
08:45
what to explore探索 and what to ignore忽視.
現在讓我給你們看看另一個例子,
08:50
Let me show顯示 you another另一個 example now,
這次是關於因果推理的。
08:52
this time about a problem問題
of causal因果 reasoning推理.
每人都要面對這個問題,
08:55
And it starts啟動 with a problem問題
of confounded混淆 evidence證據
08:57
that all of us have,
因為我們都是這世界的一部份。
08:59
which哪一個 is that we are part部分 of the world世界.
這看似不是一個問題,
但和其他問題一樣,
09:01
And this might威力 not seem似乎 like a problem問題
to you, but like most problems問題,
事情會出狀況。
09:04
it's only a problem問題 when things go wrong錯誤.
以這個嬰兒為例,
09:07
Take this baby寶寶, for instance.
所有事都出了問題,
09:09
Things are going wrong錯誤 for him.
他想開動這個玩具,但他做不到。
09:10
He would like to make
this toy玩具 go, and he can't.
我會讓你看一段幾秒的影片。
09:13
I'll show顯示 you a few-second幾秒鐘 clip.
這有兩個可能的原因,
09:21
And there's two possibilities可能性, broadly寬廣地:
可能是他做錯了一些事,
09:23
Maybe he's doing something wrong錯誤,
又或是那個玩具有些問題。
09:25
or maybe there's something
wrong錯誤 with the toy玩具.
因此在這個實驗中,
09:30
So in this next下一個 experiment實驗,
我們會給嬰兒們少許資料。
09:32
we're going to give babies嬰兒
just a tiny bit of statistical統計 data數據
這些資料會傾向支持其中一個可能性,
09:35
supporting支持 one hypothesis假設 over the other,
我們將研究這些嬰兒能否運用這些資料,
09:38
and we're going to see if babies嬰兒
can use that to make different不同 decisions決定
而作出不同的決定。
09:41
about what to do.
這個實驗是這樣的:
09:43
Here's這裡的 the setup建立.
孝媛嘗試開動那個玩具並成功了,
09:46
HyowonHyowon is going to try to make
the toy玩具 go and succeed成功.
而我的兩次嘗試都失敗了,
09:49
I am then going to try twice兩次
and fail失敗 both times,
之後孝媛再嘗試,並再次成功了。
09:52
and then HyowonHyowon is going
to try again and succeed成功,
這就像我和我的學生在使用新科技的情況。
09:55
and this roughly大致 sums總和 up my relationship關係
to my graduate畢業 students學生們
09:58
in technology技術 across橫過 the board.
重要的是這提供了少許的資料,
10:02
But the important重要 point here is
it provides提供 a little bit of evidence證據
這反映玩具並沒有問題,而是人的問題。
10:05
that the problem問題 isn't with the toy玩具,
it's with the person.
有些人可以開動這玩具,
10:08
Some people can make this toy玩具 go,
有些人則不能。
10:11
and some can't.
當這嬰兒拿到玩具時,他要作一個選擇。
10:12
Now, when the baby寶寶 gets得到 the toy玩具,
he's going to have a choice選擇.
他的母親在旁,
10:16
His mom媽媽 is right there,
所以他可以把玩具交給母親,
換另一人嘗試。
10:18
so he can go ahead and hand off the toy玩具
and change更改 the person,
同時在毛巾上有另一個玩具,
10:21
but there's also going to be
another另一個 toy玩具 at the end結束 of that cloth,
所以他也可以把玩具拉向自己,
換另一個玩具。
10:24
and he can pull the cloth towards him
and change更改 the toy玩具.
一起看看嬰兒會怎樣做。
10:28
So let's see what the baby寶寶 does.
(影片) 孝媛: 二、三、開始!
(音樂)
10:30
(Video視頻) HGHG: Two, three. Go!
(Music音樂)
羅拉: 一、二、三、開始!
10:34
LSLS: One, two, three, go!
亞瑟,讓我再試一次,
一、二、三、開始!
10:37
Arthur亞瑟, I'm going to try again.
One, two, three, go!
孝媛: 亞瑟,讓我再試吧,好嗎?
10:45
YGYG: Arthur亞瑟, let me try again, okay?
一、二、三、開始!
(音樂)
10:48
One, two, three, go!
(Music音樂)
看看這裡,記得這些玩具嗎?
10:53
Look at that. Remember記得 these toys玩具?
看到嗎? 對,我會把這個放在這裡,
10:55
See these toys玩具? Yeah, I'm going
to put this one over here,
把另一個給你。
10:58
and I'm going to give this one to you.
你拿去玩吧。
11:00
You can go ahead and play.
羅拉: 你或許會說嬰兒都愛他們的母親,
11:23
LSLS: Okay, Laura勞拉, but of course課程,
babies嬰兒 love their mommies媽媽們.
因此當玩具出現問題時,
嬰兒自然會把它交給母親。
11:27
Of course課程 babies嬰兒 give toys玩具
to their mommies媽媽們
11:30
when they can't make them work.
因此,問題在於當我們稍微改變資料時,
甚麼事會發生。
11:32
So again, the really important重要 question
is what happens發生 when we change更改
11:35
the statistical統計 data數據 ever so slightly.
這次,嬰兒將看到這玩具
按同一次序成功運作和失敗,
11:38
This time, babies嬰兒 are going to see the toy玩具
work and fail失敗 in exactly究竟 the same相同 order訂購,
但我們改變了資料的分佈。
11:42
but we're changing改變
the distribution分配 of evidence證據.
這次孝媛和我各有一次成功和一次失敗,
11:45
This time, HyowonHyowon is going to succeed成功
once一旦 and fail失敗 once一旦, and so am I.
這代表誰人嘗試都沒有分別,
那件玩具是壞的,
11:49
And this suggests提示 it doesn't matter
who tries嘗試 this toy玩具, the toy玩具 is broken破碎.
它不是每次都能運作的。
11:55
It doesn't work all the time.
同樣地,嬰兒要作出一個選擇,
11:57
Again, the baby's寶寶 going to have a choice選擇.
她的母親在旁,所以她可換另一人嘗試,
11:59
Her mom媽媽 is right next下一個 to her,
so she can change更改 the person,
同時另一個玩具就在毛巾上。
12:02
and there's going to be another另一個 toy玩具
at the end結束 of the cloth.
看看她會怎樣做。
12:05
Let's watch what she does.
(影片) 孝媛: 二、三、開始!
(音樂)
12:07
(Video視頻) HGHG: Two, three, go!
(Music音樂)
讓我再試一次,
一、二、三、開始!
12:11
Let me try one more time.
One, two, three, go!
嗯...
12:17
Hmm.
羅拉: 讓我試試吧,克拉拉。
12:19
LSLS: Let me try, Clara克拉拉.
一、二、三、開始!
12:22
One, two, three, go!
嗯...讓我再試試。
12:27
Hmm, let me try again.
一、二、三、開始!
(音樂)
12:29
One, two, three, go!
(Music音樂)
孝媛: 我把這個放在這裡,
12:35
HGHG: I'm going
to put this one over here,
這個則交給你,
12:37
and I'm going to give this one to you.
你拿去玩吧。
12:39
You can go ahead and play.
(掌聲)
12:58
(Applause掌聲)
羅拉: 看看這個實驗的結果,
13:04
LSLS: Let me show顯示 you
the experimental試驗 results結果.
在垂直軸上,你會看到在每種情況下,
嬰兒作出不同選擇的分佈。
13:07
On the vertical垂直 axis,
you'll你會 see the distribution分配
13:09
of children's兒童 choices選擇 in each condition條件,
你會發現他們作的選擇是
基於他們觀察到的資料。
13:12
and you'll你會 see that the distribution分配
of the choices選擇 children孩子 make
13:16
depends依靠 on the evidence證據 they observe.
因此當他們兩歲時,
13:19
So in the second第二 year of life,
嬰兒已經可以運用細微的資料,
13:21
babies嬰兒 can use a tiny bit
of statistical統計 data數據
在兩個完全不同的選項中作出決定:
13:24
to decide決定 between之間 two
fundamentally從根本上 different不同 strategies策略
13:27
for acting演戲 in the world世界:
尋求幫忙或自行探索。
13:29
asking for help and exploring探索.
我剛才讓你們看了兩個實驗,
13:33
I've just shown顯示 you
two laboratory實驗室 experiments實驗
在這領域中有數千個得出相同結果的實驗。
13:37
out of literally按照字面 hundreds數以百計 in the field領域
that make similar類似 points,
當中反映的重點是,
13:40
because the really critical危急 point
兒童擁有充分解讀零碎資訊的能力,
13:43
is that children's兒童 ability能力
to make rich豐富 inferences推論 from sparse data數據
這超出了所有文化的學習方式。
13:48
underliesunderlies all the species-specific種屬特異性
cultural文化 learning學習 that we do.
孩子從少數的例子便能學到新技能,
13:53
Children孩子 learn學習 about new tools工具
from just a few少數 examples例子.
他們從少數的例子便能領略到新的因果關係,
13:58
They learn學習 new causal因果 relationships關係
from just a few少數 examples例子.
他們甚至能學到新的生字,如美國手語。
14:03
They even learn學習 new words,
in this case案件 in American美國 Sign標誌 Language語言.
我會提出兩個重點作總結。
14:08
I want to close with just two points.
如果你近年有留意大腦和認知科學領域,
14:12
If you've been following以下 my world世界,
the field領域 of brain and cognitive認知 sciences科學,
14:15
for the past過去 few少數 years年份,
你會聽到三個重要的概念。
14:17
three big ideas思路 will have come
to your attention注意.
第一,現在是大腦的時代。
14:20
The first is that this is
the era時代 of the brain.
的確,神經科學近來有不少驚人的發現,
14:23
And indeed確實, there have been
staggering踉蹌 discoveries發現 in neuroscience神經科學:
例如標記了大腦皮層負責不同功能的位置、
14:27
localizing本地化 functionally功能 specialized專門
regions地區 of cortex皮質,
製造出透明的老鼠大腦、
14:30
turning車削 mouse老鼠 brains大腦 transparent透明,
以及利用光線啟動神經元。
14:33
activating激活 neurons神經元 with light.
第二個重要的概念是,
14:36
A second第二 big idea理念
現在是大數據和機器學習的時代,
14:38
is that this is the era時代 of big data數據
and machine learning學習,
而機器學習能徹底改變我們對任何事的理解,
14:43
and machine learning學習 promises許諾
to revolutionize革命化 our understanding理解
14:46
of everything from social社會 networks網絡
to epidemiology流行病學.
從社交網站到流行病學。
當機器學習能理解埸合和處理自然語言時,
14:50
And maybe, as it tackles鏟球 problems問題
of scene現場 understanding理解
14:53
and natural自然 language語言 processing處理,
也許我們能藉此了解人類的認知。
14:55
to tell us something
about human人的 cognition認識.
最後一個你會聽過的重要概念是,
14:59
And the final最後 big idea理念 you'll你會 have heard聽說
我們將對大腦有很深入的認識,
並能掌握大數據,而這很可能是件好事。
15:01
is that maybe it's a good idea理念 we're going
to know so much about brains大腦
15:05
and have so much access訪問 to big data數據,
15:06
because left to our own擁有 devices設備,
因為相比機器而言,
人類易犯錯誤,我們會走捷徑,
15:09
humans人類 are fallible易錯的, we take shortcuts快捷鍵,
我們會做錯,
15:13
we err, we make mistakes錯誤,
我們在很多方面都有偏見,
15:16
we're biased, and in innumerable無數 ways方法,
我們會有錯誤的理解。
15:20
we get the world世界 wrong錯誤.
我認為這都是重要的,
15:24
I think these are all important重要 stories故事,
因為這反映了人類的特質,
15:27
and they have a lot to tell us
about what it means手段 to be human人的,
但我今天想帶出事情的另一面。
15:31
but I want you to note注意 that today今天
I told you a very different不同 story故事.
這是關於思維而非大腦的,
15:35
It's a story故事 about minds頭腦 and not brains大腦,
尤其是人類獨有的運算能力,
15:39
and in particular特定, it's a story故事
about the kinds of computations計算
15:42
that uniquely獨特地 human人的 minds頭腦 can perform演出,
這牽涉了豐富、有條理的知識,
15:45
which哪一個 involve涉及 rich豐富, structured結構化的 knowledge知識
and the ability能力 to learn學習
以及從少量的數據和例子中學習的能力。
15:49
from small amounts of data數據,
the evidence證據 of just a few少數 examples例子.
再者,這是關於我們如何從幼童,
15:56
And fundamentally從根本上, it's a story故事
about how starting開始 as very small children孩子
一路發展至成為文化中偉大的成就,
16:00
and continuing繼續 out all the way
to the greatest最大 accomplishments成就
16:04
of our culture文化,
我們能正確地理解這個世界。
16:08
we get the world世界 right.
大家,
人腦不只是懂得從少量的數據中學習。
16:12
Folks鄉親, human人的 minds頭腦 do not only learn學習
from small amounts of data數據.
人腦能想到新的主意。
16:18
Human人的 minds頭腦 think
of altogether new ideas思路.
人腦能創造出研究和發明。
16:20
Human人的 minds頭腦 generate生成
research研究 and discovery發現,
人腦能創作藝術、文學、寫詩和戲劇。
16:23
and human人的 minds頭腦 generate生成
art藝術 and literature文學 and poetry詩歌 and theater劇院,
人腦可照顧其他人,
16:29
and human人的 minds頭腦 take care關心 of other humans人類:
包括年老的、年輕的、患病的,
16:32
our old, our young年輕, our sick生病.
我們甚至能治癒他們。
16:36
We even heal癒合 them.
在未來,我們將會看到
超乎現在能想像的科技發展,
16:39
In the years年份 to come, we're going
to see technological技術性 innovations創新
16:42
beyond anything I can even envision預見,
但在我或你們的一生中,
我們不太可能目睹比得上嬰兒運算能力的機器。
16:46
but we are very unlikely不會
16:48
to see anything even approximating逼近
the computational計算 power功率 of a human人的 child兒童
16:54
in my lifetime一生 or in yours你的.
假如我們投資在最厲害的學習者和其發展身上,
16:58
If we invest投資 in these most powerful強大
learners學習者 and their development發展,
在嬰兒和兒童身上、
17:03
in babies嬰兒 and children孩子
在母親和父親身上、
17:06
and mothers母親 and fathers父親
在照顧者和老師身上,
17:08
and caregivers護理人員 and teachers教師
如同我們投資在最厲害的科技、工程和設計上時,
17:11
the ways方法 we invest投資 in our other
most powerful強大 and elegant優雅 forms形式
17:15
of technology技術, engineering工程 and design設計,
我們不只是夢想有個更好的將來,
17:18
we will not just be dreaming做夢
of a better future未來,
而是在計劃一個更好的將來。
17:21
we will be planning規劃 for one.
謝謝。
17:23
Thank you very much.
(掌聲)
17:25
(Applause掌聲)
克里斯·安德森: 羅拉, 謝謝你,
我其實想問你一個問題。
17:29
Chris克里斯 Anderson安德森: Laura勞拉, thank you.
I do actually其實 have a question for you.
首先,這項研究真是太瘋狂了。
17:34
First of all, the research研究 is insane.
我的意思是,有誰會想到這些實驗? (笑聲)
17:36
I mean, who would design設計
an experiment實驗 like that? (Laughter笑聲)
我見過很多類似的實驗,
17:41
I've seen看到 that a couple一對 of times,
但我仍然覺得難以置信,
17:42
and I still don't honestly老老實實 believe
that that can truly be happening事件,
儘管很多人做了類似的實驗,而事實的確如此。
17:46
but other people have doneDONE
similar類似 experiments實驗; it checks檢查 out.
這些嬰兒根本是天才。
17:49
The babies嬰兒 really are that genius天才.
羅拉: 在實驗中這看似很神奇,
17:50
LSLS: You know, they look really impressive有聲有色
in our experiments實驗,
但想想在現實生活中是怎樣的,對嗎?
17:53
but think about what they
look like in real真實 life, right?
一出世時,他只是個嬰兒,
17:56
It starts啟動 out as a baby寶寶.
但18個月後他開始說話,
17:57
Eighteen十八 months個月 later後來,
it's talking to you,
而嬰兒最初說的話不只是物件,
如皮球和鴨子,
17:59
and babies'嬰兒 first words aren't just
things like balls and ducks鴨子,
他們更能表達「不見了」的概念,
18:02
they're things like "all gone走了,"
which哪一個 refer參考 to disappearance消失,
又或是以「哎喲」表達無心之失。
18:05
or "uh-oh嗯,哦," which哪一個 refer參考
to unintentional無意 actions行動.
這必須是那麼厲害的,
18:07
It has to be that powerful強大.
這必須比我剛才展示的還要厲害。
18:09
It has to be much more powerful強大
than anything I showed顯示 you.
嬰兒在弄清楚整個世界,
18:12
They're figuring盤算 out the entire整個 world世界.
一個四歲的小孩幾乎懂得說所有東西。
18:14
A four-year-old四十歲 can talk to you
about almost幾乎 anything.
(掌聲)
18:17
(Applause掌聲)
克里斯: 如果我沒錯的話,
你想指出的另一個重點是,
18:19
CACA: And if I understand理解 you right,
the other key point you're making製造 is,
這些年來,我們都聽說
我們的腦袋是不可信和會出錯的,
18:22
we've我們已經 been through通過 these years年份
where there's all this talk
18:25
of how quirky詭詐的 and buggy越野車 our minds頭腦 are,
18:27
that behavioral行為的 economics經濟學
and the whole整個 theories理論 behind背後 that
行為經濟學和其他新理論都指出我們不是理性的。
18:29
that we're not rational合理的 agents代理.
但你指出了我們的腦袋是非凡的,
18:31
You're really saying that the bigger
story故事 is how extraordinary非凡,
我們一直忽略了我們的腦袋是多麼神奇。
18:35
and there really is genius天才 there
that is underappreciated懷才不遇.
羅拉: 我最喜愛的心理學名言之一,
18:40
LSLS: One of my favorite喜愛
quotes報價 in psychology心理學
來自社會心理學家所羅門·阿希,
18:42
comes from the social社會
psychologist心理學家 Solomon所羅門 Asch阿希,
他說心理學首要的任務是
去除那些毋需証明的事物的面紗。
18:45
and he said the fundamental基本的 task任務
of psychology心理學 is to remove去掉
18:47
the veil面紗 of self-evidence自明 from things.
每天你作出大大小小的決定去理解這個世界。
18:50
There are orders命令 of magnitude大小
more decisions決定 you make every一切 day
18:55
that get the world世界 right.
18:56
You know about objects對象
and their properties性能.
你知道不同物件及其特性,
即使被覆蓋和在黑暗中你也知道。
18:58
You know them when they're occluded閉塞.
You know them in the dark黑暗.
你能在空間中行走。
19:01
You can walk步行 through通過 rooms客房.
你能猜到別人在想甚麼,你能和別人交談。
19:02
You can figure數字 out what other people
are thinking思維. You can talk to them.
你能探索空間,你明白數字。
19:06
You can navigate導航 space空間.
You know about numbers數字.
你明白因果關係,
你懂得分辨是非。
19:08
You know causal因果 relationships關係.
You know about moral道德 reasoning推理.
你毫不費力便能做到,
所以我們不會察覺,
19:11
You do this effortlessly毫不費力,
so we don't see it,
但這就是我們理解這個世界的方法,
19:14
but that is how we get the world世界 right,
and it's a remarkable卓越
這是個神奇而又難以理解的成就。
19:16
and very difficult-to-understand很難理解的
accomplishment成就.
克里斯: 我相信在坐有人認為科技正急速發展,
19:19
CACA: I suspect疑似 there are people
in the audience聽眾 who have
19:21
this view視圖 of accelerating加速
technological技術性 power功率
他們可能不認同你說電腦
不能做到三歲小孩能做到的事。
19:24
who might威力 dispute爭議 your statement聲明
that never in our lifetimes壽命
19:27
will a computer電腦 do what
a three-year-old三十歲 child兒童 can do,
但可以肯定的是,無論在甚麼場合,
19:29
but what's clear明確 is that in any scenario腳本,
嬰兒有很多地方值得我們的機器學習。
19:32
our machines have so much to learn學習
from our toddlers幼兒.
羅拉: 我同意。有些人認同機器學習。
19:38
LSLS: I think so. You'll你會 have some
machine learning學習 folks鄉親 up here.
我的意思是,你不應將嬰兒和黑猩猩跟科技比較,
19:41
I mean, you should never bet賭注
against反對 babies嬰兒 or chimpanzees黑猩猩
19:45
or technology技術 as a matter of practice實踐,
因為這不是數量上的不同,
19:49
but it's not just
a difference區別 in quantity數量,
而是性質上的不同。
19:53
it's a difference區別 in kind.
我們有十分厲害的電腦,
19:55
We have incredibly令人難以置信 powerful強大 computers電腦,
它們能做到複雜的事情,
19:57
and they do do amazingly令人驚訝
sophisticated複雜的 things,
和處理大量的資料。
20:00
often經常 with very big amounts of data數據.
我認為人類的腦袋做的事是不同的,
20:03
Human人的 minds頭腦 do, I think,
something quite相當 different不同,
人類的知識是有系統和條理分明的,
20:05
and I think it's the structured結構化的,
hierarchical分級 nature性質 of human人的 knowledge知識
這對機器仍然是一個挑戰。
20:09
that remains遺跡 a real真實 challenge挑戰.
克里斯: 勞拉·舒爾茨,十分精彩。謝謝。
20:11
CACA: Laura勞拉 Schulz舒爾茨, wonderful精彩
food餐飲 for thought. Thank you so much.
羅拉: 謝謝。
(掌聲)
20:14
LSLS: Thank you.
(Applause掌聲)
Translated by Kitty Lau
Reviewed by Ricardo Jack

▲Back to top

About the speaker:

Laura Schulz - Cognitive scientist
Developmental behavior studies spearheaded by Laura Schulz are changing our notions of how children learn.

Why you should listen

MIT Early Childhood Cognition Lab lead investigator Laura Schulz studies learning in early childhood. Her research bridges computational models of cognitive development and behavioral studies in order to understand the origins of inquiry and discovery.

Working in play labs, children’s museums, and a recently-launched citizen science website, Schultz is reshaping how we view young children’s perceptions of the world around them. Some of the surprising results of her research: before the age of four, children expect hidden causes when events happen probabilistically, use simple experiments to distinguish causal hypotheses, and trade off learning from instruction and exploration.

More profile about the speaker
Laura Schulz | Speaker | TED.com