This computer is learning to read your mind | DIY Neuroscience, a TED series

Translator: Joseph Geni Reviewer: Krystian Aparta

譯者: Joey Chung 審譯者: 至磊Zi Le 黃Ng

Greg Gage: Mind-reading. You've seen this in sci-fi movies:

葛列格·蓋奇 (GG)：讀心術。 你在科幻電影中曾看過：

machines that can read our thoughts.

可以讀出我們想法的機器。

However, there are devices today

然而，如今有很多機器

that can read the electrical activity from our brains.

可以讀出我們大腦中的電波。

We call this the EEG.

我們稱之為 「腦波圖」。

Is there information contained in these brainwaves?

這些腦波中含有資訊嗎？

And if so, could we train a computer to read our thoughts?

如果含有資訊，

我們可以訓練電腦 讀懂我們的思想嗎？

My buddy Nathan has been working to hack the EEG

我的好友內森一直 致力研究如何破解腦波圖

to build a mind-reading machine.

以建造一台可以讀心的機器。

[DIY Neuroscience]

【DIY 神經科學】

So this is how the EEG works.

先介紹一下腦波圖的原理。

Inside your head is a brain,

你的頭裡有大腦，

and that brain is made out of billions of neurons.

而大腦是由數十億個神經元構成。

Each of those neurons sends an electrical message to each other.

每個神經元都在互相傳送電子訊息。

These small messages can combine to make an electrical wave

這些微小的訊息可以結合在一起，

that we can detect on a monitor.

形成顯示器上探測到的電波。

Now traditionally, the EEG can tell us large-scale things,

傳統來說，

腦波圖能告訴我們大維度的事情，

for example if you're asleep or if you're alert.

例如你是睡著還是清醒著。

But can it tell us anything else?

但是它可以告訴我們其它事情嗎？

Can it actually read our thoughts?

它是否能夠讀出我們心中所想？

We're going to test this,

我們要測試這一點，

and we're not going to start with some complex thoughts.

而我們不會從一些複雜的想法開始。

We're going to do something very simple.

我們會從一些非常簡單的事情開始。

Can we interpret what someone is seeing using only their brainwaves?

我們只需要依據腦波 就可以判讀一個人看到了什麼嗎？

Nathan's going to begin by placing electrodes on Christy's head.

內森先在克莉絲蒂的頭上安裝電極。

Nathan: My life is tangled.

內森：我的生命亂成一團。

(Laughter)

（笑聲）

GG: And then he's going to show her a bunch of pictures

GG：之後他會給她看一些圖片，

from four different categories.

圖片來自四個不同類別。

Nathan: Face, house, scenery and weird pictures.

內森：面孔、房子、 風景、古怪的圖片。

GG: As we show Christy hundreds of these images,

GG：當我們向克莉絲蒂 展示數百張這種圖片時，

we are also capturing the electrical waves onto Nathan's computer.

內森的電腦捕捉了她的腦波。

We want to see if we can detect any visual information about the photos

我們想知道我們能否偵測

腦波中有關這些圖片的視覺資訊。

contained in the brainwaves,

so when we're done, we're going to see if the EEG

當實驗結束後，

我們將會看到腦波圖是否可以 告訴我們克莉絲蒂在看哪種圖片，

can tell us what kind of picture Christy is looking at,

and if it does, each category should trigger a different brain signal.

即是不同種類的圖片， 會否觸發不同的大腦信號。

OK, so we collected all the raw EEG data,

我們收集完了所有原始腦波圖資料，

and this is what we got.

這是我們的成果。

It all looks pretty messy, so let's arrange them by picture.

看上去很混亂， 我們來根據圖片類別分類。

Now, still a bit too noisy to see any differences,

現在，還是有點太雜亂， 無法看出任何區別，

but if we average the EEG across all image types

但是如果將圖片出現的時間對齊，

by aligning them to when the image first appeared,

並對每種類別的腦波圖取平均值，

we can remove this noise,

就能移除其中的雜亂。

and pretty soon, we can see some dominant patterns

很快我們便從各個類別中 看到一些主要的規律。

emerge for each category.

Now the signals all still look pretty similar.

現在這些信號看起來 仍然是很相似。

Let's take a closer look.

讓我們再仔細看看。

About a hundred milliseconds after the image comes on,

大約在一張圖片出現一百毫秒後，

we see a positive bump in all four cases,

我們在四個類別中 都看到了正向波動，

and we call this the P100, and what we think that is

我們把它叫作 P100 ，

我們認為這是當你識別物體時 大腦中發生的活動。

is what happens in your brain when you recognize an object.

But damn, look at that signal for the face.

但糟糕了，看看面孔圖片的信號，

It looks different than the others.

它看起來與其他的不同，

There's a negative dip about 170 milliseconds

在圖片出現後約 170 毫秒，

after the image comes on.

出現了負向波動。

What could be going on here?

這裡可能發生了什麼事？

Research shows that our brain has a lot of neurons that are dedicated

研究顯示，

我們大腦有大量神經元 專門負責識別人類的面孔，

to recognizing human faces,

so this N170 spike could be all those neurons

所以這個 N170 腦電負成分 可能是由這些神經元產生

firing at once in the same location,

在同一個地方同時啟動，

and we can detect that in the EEG.

而我們可以在腦波圖中探測到。

So there are two takeaways here.

於是這裡有兩個結論：

One, our eyes can't really detect the differences in patterns

第一，在沒有進行平均法降噪時，

我們的眼睛不能識別 腦波規律的不同；

without averaging out the noise,

and two, even after removing the noise,

第二，即使移除雜訊後，

our eyes can only pick up the signals associated with faces.

我們的眼睛也只能 識別出和面孔有關的信號。

So this is where we turn to machine learning.

於是我們在此轉而借助機器學習。

Now, our eyes are not very good at picking up patterns in noisy data,

我們的眼睛並不擅長 在雜訊中發現規律，

but machine learning algorithms are designed to do just that,

但是機器學習演算法的設計 可以解決這類問題。

so could we take a lot of pictures and a lot of data

所以我們能否將許多圖片和資料

and feed it in and train a computer

輸入到電腦中進行訓練，

to be able to interpret what Christy is looking at in real time?

從而即時判斷克莉絲蒂 究竟正在看什麼？

We're trying to code the information that's coming out of her EEG

我們嘗試將她的腦波圖資訊

in real time

進行即時編碼，

and predict what it is that her eyes are looking at.

並預測她眼睛在看什麼東西。

And if it works, what we should see

如果這樣有效，

我們應該能看到 每當她看到風景的圖片時，

is every time that she gets a picture of scenery,

it should say scenery, scenery, scenery, scenery.

機器顯示風景、風景、風景、風景，

A face -- face, face, face, face,

看到面孔── 面孔、面孔、面孔、面孔，

but it's not quite working that way, is what we're discovering.

但是我們發現，實際上並非如此。

(Laughter)

（笑聲）

OK.

好的。

Director: So what's going on here? GG: We need a new career, I think.

導演：怎麼了？ GG：我覺得我們應該轉行。

(Laughter)

（笑聲）

OK, so that was a massive failure.

好吧，所以剛剛那個是重大失敗。

But we're still curious: How far could we push this technology?

但是我們依然好奇： 我們能將這項技術推展到多遠？

And we looked back at what we did.

於是我們回顧做法。

We noticed that the data was coming into our computer very quickly,

我們發現資料飛快湧入電腦，

without any timing of when the images came on,

但沒有對圖片出現的時間進行計時，

and that's the equivalent of reading a very long sentence

這等同於讀一個 在單詞間沒有空格的長句。

without spaces between the words.

It would be hard to read,

這樣的句子很難讀懂，

but once we add the spaces, individual words appear

但是只要我們添加了空格， 我們就能看到獨立的單詞，

and it becomes a lot more understandable.

句子也就變得容易理解得多，

But what if we cheat a little bit?

但如果我們作一點弊呢？

By using a sensor, we can tell the computer when the image first appears.

透過使用感測器，

我們能告訴電腦 每張圖片出現的時間。

That way, the brainwave stops being a continuous stream of information,

這樣，腦波就不再是 一個沒有間斷的資訊串流，

and instead becomes individual packets of meaning.

而是變成了一個個有意義的封包。

Also, we're going to cheat a little bit more,

另外，我們還要再作弊一下，

by limiting the categories to two.

把圖片限制到兩個類別。

Let's see if we can do some real-time mind-reading.

讓我們看看我們是否能夠即時讀心。

In this new experiment,

在這個新實驗中，

we're going to constrict it a little bit more

我們將限制實驗條件，

so that we know the onset of the image

這樣我們就會知道圖片出現的時間，

and we're going to limit the categories to "face" or "scenery."

並將類別限制為「面孔」或「風景」。

Nathan: Face. Correct.

內森：面孔，正確。

Scenery. Correct.

風景，正確。

GG: So right now, every time the image comes on,

GG：所以現在每當圖片出現時，

we're taking a picture of the onset of the image

我們對圖片出現的時刻進行記錄，

and decoding the EEG.

並對腦波圖解碼。

It's getting correct.

它變得越來越正確。

Nathan: Yes. Face. Correct.

內森：是的，面孔，正確。

GG: So there is information in the EEG signal, which is cool.

GG：所以腦波圖的信號中 包含資訊，這很棒。

We just had to align it to the onset of the image.

我們僅僅需要把它 和圖片出現的時刻對齊。

Nathan: Scenery. Correct.

內森：風景，正確。

Face. Yeah.

面孔，沒錯。

GG: This means there is some information there,

GG：這意味著它包含了一些資訊，

so if we know at what time the picture came on,

如果我們知道圖片出現的時間，

we can tell what type of picture it was,

我們就有可能判斷 它是哪個類別的圖片，

possibly, at least on average, by looking at these evoked potentials.

至少一般可以做到， 只要根據這些由圖片誘發的電位。

Nathan: Exactly.

內森：說得沒錯。

GG: If you had told me at the beginning of this project this was possible,

GG：如果你一開始跟我說， 這個計畫有可能實現，

I would have said no way.

我會說，怎麼可能。

我真的覺得我們不可能做到。

I literally did not think we could do this.

我們的讀心術實驗真的成功了嗎？

Did our mind-reading experiment really work?

Yes, but we had to do a lot of cheating.

成功了，但是我們必須作很多弊。

It turns out you can find some interesting things in the EEG,

結果就是，你能透過腦波圖 發現一些有趣的事，

for example if you're looking at someone's face,

比如你是否在看某人的臉，

but it does have a lot of limitations.

但它確實有很多限制。

Perhaps advances in machine learning will make huge strides,

也許機器學習領域的進步 會帶來更多重大突破。

and one day we will be able to decode what's going on in our thoughts.

有朝一日，我們能夠解碼心中所想。

But for now, the next time a company says that they can harness your brainwaves

可是就現在來說，

當有公司說它能利用 你的腦波控制設備，

to be able to control devices,

it is your right, it is your duty to be skeptical.

保持懷疑是你的權利和責任。