Welcome to the World of Audio Computers | Jason Rugolo | TED

譯者: 麗玲 辛

You know, when I was a kid,

當我還是個孩子時，

dreaming about the amazing future that computers could bring,

夢想著電腦會帶來的奇妙未來，

I never thought it would look like this.

我從沒想過會變成這樣。

我在 Chipotle 墨西哥燒烤餐廳 排隊時拍了這張照片，

I snapped this photo in line at a Chipotle, thinking,

"Man, what has the world come to?"

心想，「天啊，這世界 怎麼變成這樣了？」

You know, everyone's stuck in their phones all the time.

大家總是埋頭在手機裡。

And then I almost doubled over laughing at myself because there I am,

然後我差點笑翻了， 因為我自己也在玩手機，

stuck in my phone being judgy about these people, stuck in theirs.

還批評別人玩手機。

The truth is that we could all benefit from a little less screen time.

事實上，我們都可以 少花點時間在屏幕上。

And so how do we push back

那我們如何反制

and create a healthier relationship to our technology?

並建立與科技更健康的關係呢？

I've been trying to figure out what comes next,

我一直想弄清楚接下來可能的發展，

what's the technology that we want to be using?

我們想使用什麼樣的科技？

I spent three years funding deep tech at ARPA-E,

我在 ARPA-E 花了 三年時間資助深度科技，

and then I moved to Google X,

然後轉到 Google X，

Google's moonshot factory,

谷歌的登月工廠，

before creating a spin out called iyo.

之後創建了一個 名為 iyo 的分拆公司。

The last 10 years, diligently, and some may say obsessively,

在過去的十年裡，我勤奮地， 有人可能會說是執著地，

trying to peek beyond the curve.

試圖窺探未來的趨勢。

What I think is next is that we need an entirely new kind of computer.

我認為接下來我們需要 一種全新的電腦。

One that speaks our language

一種能夠與我們交流的電腦，

instead of forcing us to speak their language of swipes and clicks.

而不是強迫我們用滑動 和點擊來與它交流的電腦。

A computer that we can talk to.

一種我們可以對話的電腦。

And not in the way that you speak at Siri

不是像對 Siri 那樣

with loud, robotic voice commands,

用大聲、機械化的語音命令，

but in the way that we talk with each other.

而是像我們彼此之間交談那樣。

So genuine, engaging conversation.

真實的、有參與感的對話。

So a new kind of natural language computing

是一種新的自然語言運算，

built on a new kind of compute hardware.

基於一種新的運算硬體。

So we've been building this new kind of computer in secret for six years now,

我們已經秘密打造 這種新型電腦六年了，

and today is the first time that we're talking about it publicly.

今天是我們第一次公開談論它。

So this is a prototype “audio computer,”

這是一個原型「語音電腦」，

as we call them.

我們是這麼稱呼它的。

And the first thing that you'll see is they don't have a screen.

您首先會注意到，它沒有屏幕。

We've got perfectly good screens in our pockets,

我們的口袋裡就有非常好的屏幕，

and we're hoping that we could leave them there or even at home

但我們希望能更常把它們放在口袋裡，

a little bit more often.

甚至留在家裡。

So these are not just fancy-looking earbuds.

這不只是看起來很高級的耳塞。

They're an entire computer,

這是一整台電腦，

and we think they're going to replace some of what we do

我們認為它將取代一些

with the visual computers that we're used to.

我們習慣用視覺電腦做的事情。

We've created a new audio-based user interface,

我們創造了一種 基於音頻的新型用戶介面，

which relies on natural language as the input

仰賴自然語言作為輸入，

and what we call “auditory space” as the output.

輸出的方式則是 我們稱之為「聽覺空間」。

And we're hoping that these innovations will help us create

我們希望這些創新能幫助我們 跟電腦建立一種新的關係，

a new kind of relationship with computers,

where the iyo is more natural, more intuitive.

語音交互技術 (iyo) 更自然、更直觀。

So you can just talk,

因此，你可以直接說話，

have a natural-language conversation

和各種不同的強大音訊應用程式

with all kinds of different powerful audio apps.

進行自然語言的對話。

So here's a really simple example.

接下來是一個非常簡單的示範。

Hey, Q.

嘿，Q。

Voice: Hi, Jason. How are you doing?

語音：嗨，傑森。你好嗎？

Jason Rugolo: Hey, I'm good, I'm doing the TED talk right now.

傑森·魯戈洛：嘿，我很好， 我現在正在進行 TED 講座。

Voice: You got this, Jason, how can I support you?

語音：你沒問題的，傑森， 我能怎麼幫你？

JR: I don't know, you know I'm nervous talking in front of this many people.

JR：我不知道，你知道 在這麼多人面前講話，我很緊張。

Voice: Yeah, I know you don't like public speaking.

語音：是啊，我知道你不喜歡 公開演講。我能怎麼幫你？

How can I help?

JR: I guess I could use a little bit of encouragement.

JR：我想我需要一點鼓勵。

Voice: Do you want me to share a positive message from a loved one?

語音：你想要我分享一條 來自親人的正面信息嗎？

JR：是的，那太好了。

JR: Yeah, that'd be great.

Maybe just one, I don't want to bore people.

也許就一條，我不想讓大家覺得無聊。

Voice: Your dad texted,

語音：你爸爸發了簡訊說，

"Good luck today, Jason. I'm so proud of you."

「祝你今天順利，傑森。 我以你為傲。」

JR: That's awesome, thanks, Q.

JR：太棒了，謝謝你，Q。

Can you tell him I said thanks and tell him I'm up on stage right now?

你能幫我回覆謝謝， 告訴他我現在在台上嗎？

He'll get a kick out of that.

他一定會覺得很有趣。

語音：沒問題。

Voice: No problem.

JR: So any suggestions for how I could loosen up a little bit?

JR：那麼，有什麼建議 可以讓我放鬆一點嗎？

Voice: Do you want to hear a joke to lighten the mood?

語音：你想聽個笑話來輕鬆一下嗎？

JR: OK, yeah, maybe if it's on theme.

JR：好啊，如果能 和主題相關就更好了。

Do you know any tech start-up jokes?

你知道有什麼科技新創公司的笑話嗎？

Voice: How many start-up founders does it take to change a light bulb?

語音：換一個燈泡 需要幾個新創公司創辦人？

JR: I don't know, tell me.

JR：我不知道，告訴我吧。

Voice: Only one.

語音助手：只需要一個。

He just holds it in place, and the world revolves around him.

他只要握住燈泡，世界就會繞著他轉。

(Laughter)

（笑聲）

JR: OK, sick burn.

JR：好吧，這真是辛辣的吐槽。

I'm going to get back to the talk here.

我得回到演講上來。

So do you see the difference between those robotic voice commands

您能聽出那些機械化語音命令

and just a conversation?

和自然對話之間的區別嗎？

Natural language is just more natural.

自然語言就是更自然。

It's intuitive, it's better.

它直觀，而且更好。

So what makes this kind of conversation possible?

那麼，這種對話是如何成為可能的呢？

It's the single most revolutionary breakthrough

這是資訊科技史上最具革命性的突破。

in the history of information technology.

And don't worry, I'm not talking about our devices,

別擔心，我不是在說我們的設備，

although they're pretty cool.

雖然它們很酷。

And I don't mean LLMs either, which are obviously a building block here.

我也不是指大型語言模型（LLM）， 儘管這顯然是其中一個組成部分。

What I'm talking about is something much older.

我說的是某個更古老的東西。

It's the first uniquely human form of communication,

那就是獨特的人類交流形式，

the one that we naturally learn as children

我們在孩提時代就自然學會了， 其結構內建在我們大腦裏。

and has its structure built into our brain.

It's the very thing I'm doing right now.

也是我現在正在做的事，

Talking.

說話。

Spoken language emerged in tandem with the evolution of human consciousness,

口語與人類意識的進化同步出現，

and to this day, it remains our most efficient

直到今天，它仍然是我們最有效

and emotionally robust form of communication.

且情感豐富的交流形式。

Conversation is not just transmitting ideas from one person to another.

對話不僅僅是將思想 從一個人傳遞給另一個人。

It's more like thinking together.

它更像是一起思考。

Modern neuroscientists have pioneered a whole new approach to the brain.

現代神經科學家開創了 一種全新的大腦研究領域，

It's called second-person neuroscience,

稱為第二人稱神經科學，

and it’s built on the notion that how we think is not isolated.

它建立在這樣一個概念上： 我們的思考方式不是孤立的，

It’s collective, and it happens out loud.

而是集體的，並且需要說出來的。

Not just through words,

不僅通過語言，

but through subtle signals of tone and prosody,

還通過語調和韻律、

your timbre and your pitch and intensity.

音色、音高和強度的微妙信號。

And neuroscience is just not complete

除非你將第二個人加入 這個完整的社會動態中，

until you add a second person into this full social dynamic.

否則神經科學是不完整的。

So why can't we have a computer that we can talk with in that way?

所以，為什麼我們不能有

一台可以用那種自然語言 進行對話的電腦呢？

With that kind of natural language.

A computer that has superhuman processing speed.

一台具有超人處理速度的電腦，

And it has access to the internet.

能連接網路，

And it’s been trained on the entire written record of human thought.

而且已經學習了人類思想的 所有書面記錄。

But engages with you like a person would,

但它能像一個人一樣與你交流，

that understands your intention

理解你的意圖，

and that taps into the superpower of human natural language understanding.

並利用人類理解自然語言的超能力。

That's the promise of audio computing.

這就是語音運算的承諾。

So think about not just how it can replace many of the things

想想，它不僅可以取代

that you do on your phone

你在手機上做的許多事情，

but actually make them better.

而且能做得更好。

So take email, for example.

以電子郵件為例。

We pull out our phones, we swipe, we scroll,

我們掏出手機，左右滑、往下滑，

we furiously type with our thumbs.

用拇指拼命打字。

Wouldn't it be better to just sit back with a cup of coffee

假如能坐下來喝杯咖啡，

一面談話一面聽取簡報，不是更好嗎？

and to be briefed in a conversation?

Or search, search is a big one.

再說搜尋，搜尋是件大事。

It's an incredible technology

這是一項令人難以置信的科技，

that made the world a radically better place.

它使世界變得更加美好。

But with these audio computers,

但是有了這些語音電腦，

you can just talk out loud about anything that you want to know.

你只需要大聲說出 你想知道的任何事情。

It just feels so normal.

這感覺如此自然。

So there's a big difference between giving a voice command

這與向五大語音助手之一

to one of the big five voice assistants,

發出語音命令有很大不同，

which are these structured, predefined choose-your-own-adventure dialogue models

那些是結構化、預定義的 「選擇你自己的冒險」對話模型，

that we all have to learn,

我們都必須學習才能用，

and I'm sure have all felt that frustration,

而且我相信我們都曾感到挫敗，

and just having a real conversation.

而與此相對的則是進行真正的對話。

These natural-language applications can get to know you

這些自然語言應用程序能夠了解你， 就像我們彼此了解一樣。

in the same way that we get to know each other.

They build context about our lives just through us talking over time.

它們透過和我們交談 漸漸建立我們生活的情境。

So later, take out your phone, look at all those apps,

稍後，請拿出你的手機， 看看那些應用程序，

all those candy-colored icons,

那些糖果色的圖標，

and think about how could you accomplish the same thing

想一想你如何通過對話 來完成同樣的事情？

but through conversation?

Or how could you make it better?

或者如何使過程變得更好？

You won't be able to do Instagram or TikTok,

你無法使用 Instagram 或 TikTok，

those apps whose content is mostly visual.

那些應用的內容主要是視覺的。

But wouldn't it be better to spend a little bit less time in those apps,

但減少花在這些應用程式上的時間，

or just to need your screen a little bit less?

或者減少對螢幕的需求不是更好嗎？

So our goal is to be heads-up and hands-free

我們的目標是一天中

for a little bit more of the day.

更多時間抬起頭來，解放雙手。

You know, just get back into the world.

重新回到世界中去。

Of course, if the auditory user interface or the AUI, as we call it,

當然，如果語音用戶界面 （我們稱之為 AUI）

is going to really integrate into your life,

真的要融入你的生活，

it has to feel private and convenient to use.

它必須感覺私密且使用方便。

So that's why we built it as an all-day wearable for the ear.

因此，我們將其設計為 全天佩戴的耳掛式設備。

But your ears are for hearing first and foremost.

但是你的耳朵首先是用來聽的。

And so if you're going to wear a computer on them all day,

所以，如果你要整天 佩戴一台電腦在耳朵上，

we can't mess that up.

我們不能搞砸了。

In fact, we should probably make that better too.

事實上，我們應該讓這變得更好。

So these audio computers,

因此，這些語音電腦，

over the last six grueling years of R and D,

在過去六年艱苦的研發中，

became a sort of mixed-reality device.

變成了一種混合實境設備。

It's like the Apple Vision Pro, but for audio,

就像 Apple Vision Pro， 但這是針對音頻的，

where we can pass through and we can modify your ambient acoustics,

我們可以通過它傳遞 和修改你的環境聲音，

giving you an unprecedented control over your personal soundscape.

給你前所未有的個人音景控制能力。

In order to do mixed-audio reality,

為了達成混合音頻實境，

we sort of had to hack the auditory system

我們不得不破解聽覺系統，

to be able to display sound in ultra-high fidelity, spatially,

以超高保真度和空間感來呈現聲音，

as if it's all around you.

就像它環繞在你周圍一樣。

So there's this whole field of research.

有這麼一個研究領域，叫做心理聲學，

It’s called psychoacoustics, which we’ve led on for years.

我們在這方面領先多年。

We built this giant audio structure.

我們建造了一個巨大的音頻結構。

It's a dome with 128 custom speakers coming from all directions,

它是一個圓頂，有 128 個 定制揚聲器從各個方向傳來，

so we could create virtual soundscapes.

如此，我們可以創建虛擬音景。

It's sort of like the Star Trek holodeck, but for audio.

有點像《星際迷航》的全息甲板， 但這是針對音頻的。

And if you're standing in the middle of this and you close your eyes,

如果你站在這個結構的 中間，閉上眼睛，

we can transport you auditorily to anywhere that we want.

我們可以在聽覺上 將你傳送到任意一個地方。

So we can render a virtual football game,

我們可以呈現一場虛擬的足球比賽，

and you feel the energy.

你能感受到那脈動。

Or we can make it sound like you’re in the middle of a bustling city street.

或者我們可以讓它聽起來 像你在繁華的城市街道中。

And if you’re me, you feel the anxiety.

如果你像我，你就會感到焦慮。

Or standing on a beach with the crashing waves,

或者站在有海浪聲的海灘上，

and you feel the peace.

你會感到平靜。

And so it's super cool.

這真的很酷。

I wish everyone could be inside there.

我希望每個人都能親自體驗一下。

Then we ran countless experiments to figure out all the complicated ways

我們進行了無數的實驗

來找出大腦在空間中 定位聲音的所有複雜方式。

that your brain positions sounds in space.

Also, we could reverse-engineer those neural algorithms

我們還對這些神經演算法 進行逆向工程，

and code them into our software.

並將其編碼到我們的軟體中。

So our goal has been to create this experience

我們的目標是創造這種體驗，

but right here.

但就用這個設備。

Us psychoacousticians call this “virtual auditory space”

我們心理聲學家稱之為 「虛擬聽覺空間」，

to distinguish from our real auditory space,

以區別於我們的真實聽覺空間，

which is, you know, the sounds that are all around us.

也就是我們周圍的聲音。

And this is what's necessary to create a compelling mixed audio-reality device.

這是創建一個引人入勝的 混合音頻實境設備必需做到的。

So it's actually impossible to demonstrate this experience

除非您親耳聽到，

until you hear it with your ears yourself.

實際上不可能展示這種體驗，

But to give you an idea, we have tried to simulate it for you.

但為了讓您有概念， 我們嘗試在此模擬一下。

So imagine that you’re sitting in a noisy restaurant,

想像一下您坐在一個嘈雜的餐廳裡，

and you're having trouble hearing your friends across the table.

難以聽清對面朋友的說話聲。

(Overlapping voices, music and noise)

（重疊的聲音、音樂和噪音）

Hey, can you enhance the sounds that are right in front of me?

嘿，你能增強正前方的聲音嗎？

(People speaking)

（人聲交談）

(Baby crying)

（嬰兒哭聲）

And can you turn that baby down?

能把那個嬰兒的聲音調低嗎？

(People talking)

（人聲交談）

That’s better.

這樣好多了。

I'm still having a little trouble hearing Pedro.

我還是有點難聽清佩德羅的聲音。

Can you isolate Pedro for me?

能幫我隔離佩德羅的聲音嗎？

Pedro: (Speaking in Spanish)

佩德羅：（用西班牙語講話）

JR: That's perfect.

JR：這很完美。

And, you know, my Spanish is a little rusty.

你知道，我的西班牙語有點生疏。

Can I hear Pedro but in English?

我可以聽佩德羅的聲音， 但轉換成英語嗎？

Pedro: And at the end of the trip,

佩德羅：旅行結束時，

we came back to the city to visit the historic center.

我們回到了城市，參觀了歷史中心。

JR: Hey Shell, close all programs.

JR：嘿，Shell，關閉所有程式。

(Noise enhances)

（噪音增強）

Ah, it’s so much worse.

啊，糟糕多了。

That's pretty cool, right?

這很酷，對吧？真的很酷。

It's pretty cool.

真的很酷。

(Applause)

（掌聲）

So what you just heard was a beamforming app,

您剛剛聽到的是 一個波束成形應用程式、

the computational auditory scene analysis app,

一個聽覺場景演算分析應用程式、

a machine-learning denoising app,

一個機器學習降噪應用程式、

an AI transcription and translation

一個 AI 轉錄和翻譯

and text-to-speech with style transfer app.

及帶風格轉換的文本轉語音應用程式。

The point is that all those audio transformations are done by software.

關鍵是所有這些音訊轉換 都是由軟體完成的。

So we think the possibilities for these audio computers

所以我們認為這些語音電腦的可能性

are pretty much endless,

幾乎是無限的，

and we can't wait to see

我們迫不及待地想看看

what the world's developers are going to do here.

全球的開發者在這部份能做什麼。

Like imagine an education app that knows your personal learning style

比如想像一下，了解你 個人學習風格的教育應用程式，

and can teach you with the quality of a world-class professor,

可以隨時以世界級教授的質量教你。

on-call anytime.

Or like a fitness coach you can summon all day

或者想像一個你可以 隨時召喚的健身教練，

about your diet and exercise,

照顧你的飲食和運動，

who can also motivate you through conversation

還可以通過對話激勵你，

and even gamify your workout with some auditory cues.

甚至透過一些語音提示 將你的鍛煉遊戲化。

Or, hey, K?

或者，嘿，K？

Voice: Hi, Jason. What's up?

語音：嗨，傑森，怎麼了？

Hey, if you were going to make an audio app that could be anything,

R：嘿，如果你要做一個 音頻應用，什麼都可以，

what would it be?

你會做什麼？

Voice: How about a whoopee cushion

語音：放屁坐墊怎麼樣？ 每次你坐下來，就會發出放屁聲音。

that plays a fart sound whenever you sit down?

(Laughter)

（笑聲）

JR: Hey, K, if you were going to make an audio app

JR：嘿，K，如果你要做一個 與放屁無關的音頻應用，

that didn't have anything to do with farts, what would it be?

你會做什麼？

Voice: Maybe an app that generates personalized soundscapes

語音：也許是個生成個人化聲景的 應用程式，幫助放鬆與專注？

for relaxation and focus?

JR: That's much better.

R：那好多了。

Alright, it looks like we still have a little fine-tuning to do here.

好吧，看來我們還需要做一些微調。

So the point is,

重點是，

imagination is the only limit to what you can do here.

在這方面的未來發展， 想像力是你的唯一限制。

Our goal is not just to create the world's first audio computer,

我們的目標不僅僅是 創造世界上第一台語音電腦，

it's to create a truly intuitive computing experience

而是創造一種真正直觀的演算體驗，

where we're not monetizing your attention

我們不會利用你的注意力賺錢，

or making you captive to a new kind of device,

或者讓你成為某種新型設備的俘虜，

but instead interfacing machines with us

而是以我們與生俱來的方式 來與機器互動。

in the way that we were born to.

So I think it's time for a computer that speaks our language.

我認為該是時候 讓電腦說我們的語言了。

Thank you.

謝謝。

(Applause)

（掌聲）