The Mind-Reading Potential of AI | Chin-Teng Lin | TED

翻译人员: Yip Yan Yeung 校对人员: Yanyan Hong

How often are you frustrated by the time it takes

你有多常因为把时间花在

to accurately get things in your mind into a computer?

准确地把脑海里的东西 输进电脑感到抓狂？

It is even worse for people like me,

对我这样的人来说甚至更糟，

whose first language is not based on letters.

我们的母语不是基于字母。

I live and work in Australia,

我在澳大利亚生活和工作，

but I am originally from Taiwan.

但我来自台湾。

I moved to Sydney eight years ago

我八年前移居悉尼，

and now run a university research center there.

现在在悉尼经营一所大学研究中心。

Most of us use keyboards every day

大多数人每天都使用键盘

to get things in our minds into the computer.

将脑海中的东西输入计算机。

We have to learn to type.

我们得学会打字。

The fact that you have to learn to do some things

你必须“学会”做什么事，

shows how unnatural it is.

说明有多么违反天性。

The finger-driven touch screen has been around for 60 years.

由手指驱动的触摸屏 已经存在了 60 年。

It's convenient, but it is also slow.

很方便，但也很慢。

There are other ways to control computers --

还有其他控制计算机的方法——

joystick or gestures --

操纵杆或手势——

but they are not very useful in capturing the words in your mind.

但它们不是很擅长 捕捉你脑海中的文字。

And it is words --

文字对于人类的交流至关重要。

they are critical to communication for human beings.

The problem is about to be over,

这个问题很快就会得到解决，

because of AI.

就是因为 AI。

Today, I will show you

今天，我将向你展示

how AI can turn the speech in your mind into words on screen.

AI 如何将你脑海中的语音 转换成屏幕上的文字。

Getting from the brain to the computer efficiently

高效地从大脑传到计算机

is a real bottleneck for any computer application.

是任何计算机应用程序真正的瓶颈。

It has been my passion for 25 years.

它是我 25 年来 一直乐于研究的话题、

Many of you, or most of you,

很多人，或者大多数人，

have heard of "brain-computer interface," BCI.

都听说过“脑机接口”，即 BCI。

I have been working on BCI,

我一直在研究 BCI，

for the direct communication between the brain and machine,

用于大脑和机器之间的直接通信，

since 2004.

从 2004 年开始。

I developed a series of EEG headsets that do this.

我开发了一系列脑电波头戴设备 完成这一任务。

But they are not new.

但它们本身不新鲜。

What is new is an interface that works in a natural way,

新鲜的是自然的交互方式，

based on how our brain is working naturally.

基于我们大脑的自然运行方式。

Imagine reading words when someone is thinking,

想象一下， 当有人思考时阅读单词，

translating the brain signals into words.

将大脑信号转化为文字。

Today, you will see this in action,

今天，你们会亲眼见证，

and with no implants.

而且无需植入。

We are using AI

我们使用 AI 解码你脑中的大脑信号，

to decode the brain signals on the top of your head

and identify the biomarkers of speaking.

并识别说话的生物标记。

That means that you can send the words in your mind into the computer

意味着你可以将脑海中的文字 传输到电脑里，

with wearable technology.

借助可穿戴式设备。

It's exciting,

振奋人心，

and I believe it will open up the bottleneck

我相信它将突破

of how we engage with computers.

我们与计算机互动的瓶颈。

We are making exciting progress in decoding EEG to test this.

我们在解码脑电波上取得了显著进展， 让我们可以进行测试。

It's natural.

这很自然。

We have had very promising results

我们在解码人们说出声时的脑电波方面

in decoding EEG when someone is speaking aloud.

得到了很积极的结果。

The frontier we are working on now

我们现在正在研究的前沿领域

is to decode EEG when the speech is not spoken aloud,

是解码不说出声时的脑电波，

the words that flow in your mind when you are listening to others

当你聆听他人

或自言自语、自己在思考时 脑海中浮现的文字。

or when you are talking to yourself or thinking.

We are well on the way to make it a reality.

我们正在将其实现。

Who would like to see this in action?

谁想亲眼看一下？

(Cheers and applause)

（欢呼和掌声）

Great, we are ready to demonstrate it to you.

太棒了，我们已经准备好 向大家演示了。

I am going to invite two of my team, Charles and Daniel,

我将邀请我的两位团队成员， 查尔斯和丹尼尔，

to show it to us again.

再次呈现给我们看。

This is the first world premiere for us,

这是我们的首次全球首演，

so I hope you can be patient with us.

所以敬请保持耐心。

We are getting around 50 percent accuracy ...

我们的准确率约为 50％……

(Laughter)

（笑声）

in decoding the brain signals into words

将大脑信号解码成文字，

when someone is speaking silently.

在人们无声地“说话”时。

Here shows how it will work.

这就是它运作方式。

We have a collection of words that we have trained our technology with.

我们有一系列的词汇 用于训练我们的技术。

They are combined into sentences.

它们被组合成句子。

Charles will select one sentence,

查尔斯将选择一句话，

and Daniel will read the sentence word by word, silently,

丹尼尔将默默地逐字读出句子

and produce the brain signals that will be picked up by our sensors.

并产生大脑信号， 由我们的传感器采集信号。

Our technology will decode the brain signals into words.

我们的技术会将大脑信号解码成文字。

Charles, Daniel, are you ready to go ahead?

查尔斯，丹尼尔， 准备好开始了吗？

This is the sentence

这是丹尼尔要默读的句子。

that Daniel is going to read silently.

(Applause)

（掌声）

Sorry, please keep silent. (Laughter)

不好意思， 请保持安静。（笑声）

Here are the --

这是……

decoded words.

解码得出的单词。

They are likely the intended words.

它们可能是预期中的单词。

You can see the probability ranking

你可以看到概率排名，

of the decoded words by our technology.

由我们的技术罗列解码出来的单词。

The pattern shows our predicted sentence ...

这个结果展示了我们预测的句子

is not so correct.

不是那么正确。

(Laughter)

（笑声）

Sorry, you see the other 50 percent,

对不起， 你看到了失败的 50%，

the system doesn't work.

系统运行失败了。

But actually, you still can see some keywords where we got it.

但其实你仍然可以看到 一些我们得出的关键字。

Let's invite Charles and Daniel to do it again,

让查尔斯和丹尼尔再来一次，

but please keep silent when he's reading silently.

但是当他默默阅读时，请保持安静。

Here is the sentence,

这句话，

another sentence that Daniel will read word by word, silently.

丹尼尔逐字阅读这一句，默读。

(Laughs)

（笑声）

Again, here are the decoded words.

这是解码后的文字。

They are likely the intended words.

它们可能是预期中的词语。

The pattern shows our predictive sentence

结果表明我们预测的句子

is very close to the ground-truth sentence this time.

与原始的句子非常接近。

(Cheers and applause)

（欢呼和掌声）

Thank you, thank you.

谢谢，谢谢。

How does it work?

它是如何运作的？

We pick up the brain signals with sensors

我们用传感器采集大脑信号，

and amplify and filter them

然后对其进行放大和过滤，

to reduce the noise and get the right biomarkers.

为了减少噪音 并获得正确的生物标志。

We use AI for the task.

我们使用 AI 来完成任务。

We use deep learning to decode the brain signals

我们使用深度学习将大脑信号 解码成预期的单词。

into the intended words.

And then we use the large language model

然后我们使用大语言模型

to make the match of the decoded words

匹配解码后的单词，

and make up for the mistakes in EEG decoding.

弥补脑电波解码中的差错。

All of this is going on in the AI,

这一切都是在使用 AI，

but for the user, the interaction is natural,

但对于用户来说，通过思想 和自然语言进行互动是自然的。

through thoughts and natural language.

We are very excited about the advances that we are making

我们对我们在理解单词和句子方面

in understanding words and sentences.

取得的进步感到非常兴奋。

Another thing that is very natural to people

人们感到很自然的另一点是

is looking at something that has their attention.

看着引起他们注意的东西。

Imagine if you could select an item just by looking at it,

想象一下，如果你只需要 看着一件物品就能选择它，

not by picking it off the shelf

而不是从货架上把它拿下来

or punching a code into the vending machine.

或在自动贩卖机上输入一个号码。

Two years ago,

两年前，

in a project about hands-free control of robots,

在一个不用手操作的 机器人控制项目中，

we were very excited about robot control

我们感到非常兴奋，实现了

via visual identification of the fingers.

利用对手指的视觉识别 完成了机器人控制。

We are now beyond that.

我们现在已经不止于此。

We need not any fingers.

我们不需要任何手指。

The AI is making it natural.

AI 让它更自然。

There are four objects on the table.

桌子上有四个物体。

Toy car,

玩具车、

toy animal,

玩具动物、

plastic flower and some food,

塑料花和一些食物，

which is also plastic,

也是塑料的，

not left over from the breakfast this morning.

不是今天早餐剩下的。

You can also see the four objects' photo on the screen.

你也可以在屏幕上 看到这四个物体的照片。

Daniel is going to look at the photos,

丹尼尔会看着照片，

and select an item in his mind.

然后在脑海中选择一件物品。

If it is working as it should,

如果它能正常工作，

you will see the selected item pop up on screen.

你们会看到屏幕上跳出选中的物品。

We use photos for this because they are very controllable.

我们使用照片， 是因为它们非常易于控制。

To show that this is not all just built into my presentation,

为了证明这不是我演讲 预设的一部分，

Charles will pick up one item for Daniel to select in mind.

查尔斯会为丹尼尔选择一个物品， 让他在脑海中选择。

Please, Charles.

请吧，查尔斯。

It's a car.

这是一辆车。

So, Daniel, select ...

那么丹尼尔选中了……

the car in his mind.

在脑海中选择了车。

(Laughter)

（笑声）

Hamburger.

汉堡。

(Laughter)

（笑声）

It's incorrect.

这是不正确的。

(Laughter)

（笑声）

It's unlucky that the 30-percent error rate came with us again.

不幸的是，我们又出现了 30% 的错误率。

Let's invite Charles and Daniel to show it again.

让查尔斯和丹尼尔再演示一次。

It's a duck, a lovely duck.

这是一只鸭子，一只可爱的鸭子。

(Laughter)

（笑声）

OK. Good.

好吧。很好。

(Cheers and applause)

（欢呼和掌声）

Thank you. Thank you.

谢谢。谢谢。

Daniel did this for his PhD.

这是丹尼尔的博士课题。

It's very impressive, isn't it?

非常了不起，不是吗？

When Daniel selects an item in his mind,

当丹尼尔在脑海中选择一件物品时，

his brain recognizes and identifies the object and triggers his EEGs.

他的大脑会认出、识别 该物体并触发他的脑电波。

Our technology decodes the triggers.

我们的技术可以解码触发因素。

We are working on our way through the technical challenges.

我们正在努力攻克技术挑战。

We will work on overcoming the interference issue.

我们会努力克服干扰问题。

That's why I asked for the phones to be turned off.

这就是为什么我请大家关掉手机。

Different people have different neural signatures,

不同的人有不同的神经特征，

which are important to decoding accuracy.

这对于解码的准确性很重要。

One reason I brought Daniel along here

我邀请丹尼尔来到这里的原因之一

is because he can give off great neural signatures.

是因为他能发出很好的神经特征。

(Laughter)

（笑声）

(Applause)

（掌声）

Yeah, he can give us a great neural signature,

是的，就我们的技术而言， 他能给我们很好的神经特征。

as far as our technology is concerned.

They are still cables here as well.

还是会需要电缆。

It is not yet very portable.

还不是很便携。

Probably one biggest barrier to people using this

人们使用它的最大障碍可能是：

will be: “How do I turn it off?”

“我怎么把它关了呢？”

Any one of you will have had times when you are happy

有没有人都有过这样的经历，

that the people you are with don't know what you are really thinking.

就是你旁边的人都不知道 你到底在想什么。

(Laughter)

（笑声）

There are serious privacy and ethical issues

有严重的隐私和道德问题需要解决。

that will have to be dealt with.

I am very passionate about how important this technology can be.

我对这项技术的重要性充满热情。

One exciting point is linking the brain-computer interface

一个令人兴奋的地方是将脑机接口

to wearable computers.

与可穿戴计算机连接起来。

You already have a computer on your head.

你的头上已经有一台电脑了。

The brain will be a natural interface.

大脑将是一个自然的接口。

It is not only about controlling a computer.

这不仅仅是控制计算机。

The natural BCI also provides

自然的 BCI 还为人们提供了

another way for people to communicate with people.

另一种与人沟通的方式。

For example, it allows people who are not able to speak

例如，它让无法说话的人

to communicate with others,

与别人交流，

or such as when privacy or silence are required.

或者在需要隐私或沉默时。

If your idea of nature is a lovely forest,

如果你认为大自然 是一片美丽的森林，

you could wonder how natural this could be.

你可能在想它会有多么自然。

My answer is, it's natural language,

我的答案是，这是自然语言，

it's the natural thought process that you are using.

是你正在使用的自然思维过程。

There are no unnatural implants in your body.

你体内没有不自然的植入物。

I am challenging you to think

我想请你思考一下

about what you regard as natural communication.

你所认为的自然沟通是什么。

Turning the speech in your mind into words.

把你脑海中的言语变成文字。

There is a standard way to finish up when talking with people --

在与人交谈时， 有一种标准的结束方式——

you say: “Just think about it.”

你会说：“想一想吧。”

I hope you are as excited as we are

我希望你和我们一样

for the prospect of a future

对未来的前景感到兴奋，

in which, when you just think about something,

在这个未来中，当你思考某件事时，

the words in your mind appear on screen.

你脑海中的文字就会出现在屏幕上。

Thank you.

谢谢。

(Applause)

（掌声）