So, the one that caught my eye here is Pinterest, which⏤we all know and love Pinterest.

吸引我注意的是 Pinterest 。我們都知道而且愛用 Pinterest 這個網站。

"Determine the topical relatedness between pieces of text", 40 cents.

「找出文字段落中的主題關聯性。」40 分錢。

Wait, there's one more⏤so, this one's "find info from an email" for 3 cents.

等等，還有更多，這個是「從電子郵件中找出資訊。」3 分錢。

They're just gonna email me, and then I, like, find info from it?

他們就寄電子郵件給我，然後我從裡面找指定資訊就好嗎？

Which seems cool.

好像蠻容易的。

It turns out, Amazon runs an online marketplace that farms out basic tasks computer programs have a hard time with.

結果 Amazon 有經營一個把電腦程式無法處理的基本作業外包出去的網路市場。

It's called Mechanical Turk, named after a robot from the 1700's.

它叫作 Mechanical Turk (機械土耳其人) ，命名自西元 1700 年代的一個機器人。

But people just call it MTurk.

但人們都稱它作 MTurk。

So, what I wanna do now is, Leonard Monteiro will pay me 3 cents to write the prices shown in an image.

所以我現在要來作的就是，寫下圖片中的價錢，Leonard Monteiro 就會付我 3 分錢。

I'm not totally sure why this is, like, worth-doing money, but, okay, let's do it.

我不太確定為什麼這個值得花錢做，但好吧，就來試試看。

It looks like a parking meter?

這看起來...這是停車計時器嗎？

I hope that this is for a good cause somehow, that this is, like, this app is helping people rather than just sending bills to people.

我希望這是出自於某種好理由，這個應用程式是在幫助人而非只是寄帳單出去。

So, I feel like on the other end of this task, there's some sort of automated parking meter app, which is weird.

我覺得在這個作業的另一端，有某種自動停車計時器的應用程式，感覺很怪。

But it turns out, stuff like this happens all the time.

但其實像這樣的事情常常發生。

Nearly every successful AI product has human beings behind it.

幾乎每個成功的人工智慧產品背後都有人類在作業。

You just don't see them until you look at the big picture.

你只是在看見全貌前沒有發現而已。

So, what's going on here?

所以這到底是怎麼回事？

Is there actually an app that claims to read parking meter fines, but it's actually humans doing it?

會不會其實那些宣稱能判讀停車計時器費用的應用程式都是人類在辨識的？

Or, am I helping train their AI and these meters are just hypothetical examples?

還是我是在協助訓練他們的人工智慧系統，然後這些只是假設性的範例？

To figure that out, I asked a resident AI expert James Vincent.

為了要找到解答，我詢問了人工智慧的常駐專家 James Vincent 。

So, I did a little Googling here, and, yeah, sorry, Russell, this is not a parking meter at all.

我查了一下，抱歉 Russell ，這個不是停車計時器。

This is actually a little gadget you put in supermarkets and you scan barcodes to check the prices.

這其實是放在超市拿來掃描條碼查看價錢的工具。

Now, the bigger question is: Why does someone want you to write down all these prices?

現在呢，為什麼有人要付錢讓你寫下這全部的價錢呢？

And I have two answers for you.

我要告訴你兩個解答。

In the first case, creating training data.

在第一個情況下，是要建立訓練資料。

Say you want to make a machine vision system that automatically does what you're doing.

假設你要製造一個自動完成你正在做的事的機器視覺系統。

How does it actually know where to look in the picture?

它該如何知道要看向圖片中的哪裡？

How does it know what the⏤what this barcode scanner looks like?

它要怎麼知道條碼掃描機看起來怎樣？

To teach it that information, you need to feed it labeled data.

為了要教導它那個資訊，你必須提供它標籤化的資料。

You need to get a human to do that labeling, in this case, Russell.

你需要一個人類來做標籤，而在那個狀況下，就是 Russell 。

He labels the data, it goes into the system, and the system learns what these things look like.

他來替這些資料做標籤，然後進到系統裡，系統就會學習並記下這些東西的樣子。

Oh, there's so much more!

噢不，還有更多其他的要打！

That is how you train an AI system.

這就是訓練人工智慧系統的方法。

But, sometimes, these systems, they don't work, right?

但有時候這些系統沒有用，對不對？

So, you use case number two⏤what's that?

所以就來到了第二個情況，是什麼呢？

Well, that might be where the AI system actually can't do what it says it can do.

那可能就是人工智慧系統其實沒辦法做他應該要做到的事。

It might be that there's too much glare in the picture and it can't read the numbers on the screen very well.

可能是圖片反光太嚴重導至它沒辦法好好辨識出螢幕上的數字。

In those cases, you need to throw the data to someone who has the intelligence to work out what's going on.

在那些情況下，你必須把資料丟給有智慧解決這些問題的人。

That's not a machine, that's a human, like Russell.

不是一個機器，而是像 Russell 一樣的人類。

And they will label the data for the machine and return it back to the end user.

他們會代替機器來做標籤，然後回傳給末端使用者。

Sometimes companies are upfront about this, and sometimes they lie about it, too.

有些公司對這個相當坦誠，而有些公司也會說謊隱瞞。

Sometimes, they will say, "Yes, we've got a whizzy AI system that's doing all this automatically."

有時候他們會說「沒錯！我們有個聰明的人工智慧系統正自動完成所有事情。」

And, actually, they don't.

而事實上，並非如此。

Turns out that that AI is a lot of low-paid workers on a system like Mechanical Turk, like Russell, providing this data in the background.

結果人工智慧只是一堆，像 Russell 一樣，在 Mechanical Turk 這種系統上的廉價勞工在後台提供資料。

If you've ever filled out a CAPTCHA, you've probably done some of that work yourself.

如果你曾經填寫過驗證碼，你自己可能就曾做過那些工作。

In theory, those tests are meant to verify that you're human, but Google has started using them to collect data for other products, too.

理論上這些測驗本來是為了要證實你是人類，但 Google 已經開始用它們來為其他產品蒐集資料。

Typing out this blurry word could help the character recognition algorithm in Google Books.

打出這個模糊的字可以幫助 Google Books 的字形辨識演算法。

These skewed numbers are probably helping confirm an address in Google Street View.

這些扭曲的數字可能可以幫助 Google 街景服務確認地址。

The most recent CAPTCHAs ask you to identify all the squares of a picture that have a car in it, at the same time, the Google's Waymo branch is trying to train self-driving cars.

最新的驗證方式是要你辨識出有汽車在內的圖片，同一時間 Google 的 Waymo 分支正嘗試訓練自動駕駛汽車。

Even a simple task like setting a timer with Google Assistant can require an army of contractors manually annotating the data, as a recent "Guardian" investigation showed.

近期英國衛報的研究顯示，甚至像是用 Google 助理設鬧鐘這麼簡單的一個作業，都需要龐大的承包者人工註解那些資料。

Sometimes, users do the labeling themselves.

有時候使用者自己做了那些標籤。

Facebook has some of the best facial recognition data in the world because they already have dozens of pictures of your face.

臉書擁有世界上最佳的人臉辨識資料之一，因為他們早就有好幾張你的臉的照片。

You added them yourself.

是你自己加上去的。

Multiply that across billions of users, and it's all the data you need to build a facial recognition system, which can then start automatically tagging your friends in the next set of pictures you upload.

把數十億的使用者數據加起來，就足以建造出在下次上傳照片時自動標記好友的人臉辨識系統。

Suddenly, Facebook has one of the most advanced facial recognition systems in the world, and they didn't have to pay a dime for it.

突然間，臉書就有了全世界最先進的人臉辨識系統之一，而且他們根本不用花半毛錢。

When researchers at Google were trying to build a depth-sensing camera, they went even further.

當 Google 的研究人員嘗試要建造一個深度感測相機，他們更進了一步。

What they really needed were a bunch of videos where mobile cameras explored static space from different angles.

他們真正需要的是一堆手機相機從不同角度探索靜態空間的影片。

But where would they find that?

但他們要從哪裡找這些影片呢？

Google downloaded 2,000 mannequin challenge videos, fed them into an algorithm, and a new kind of depth-sensing software was born.

Google 下載了 2000 部假人挑戰影片，把它們放進演算法中，一個新的深度感測軟體就誕生了。

Think about it: Every minute, 500 new hours of content are added to YouTube.

想想看， 每分鐘有 500 小時長的全新內容被加到 YouTube 上面。

If you're training an AI, that's a lot of video to draw on.

如果你要訓練一個人工智慧，有非常多的影片可以取材。

And there are no copyright restrictions on what you can use for training data.

而且用來做訓練的資料並沒有任何版權限制。

The same goes for websites, images, Wikipedia pages⏤it's all just there for the taking.

網站、影像、維基百科頁面也一樣唾手可得。

This has been a huge driving force for the AI boom.

這一直是造成人工智慧迅速發展巨大的推動力。

These systems need lots of examples to recognize even the most basic patterns.

甚至在辨識最基本的圖形，這些系統也需要非常多的範例。

That used to mean months of data entry, but now you can scrape everything you need from the internet in a matter of hours.

那曾經需要輸入好幾個月的資料，但現在只要花幾小時從網路上獲取所有你需要的資料就可以了。

And the people who made the mannequin challenge videos, they didn't think they were encoding depth information.

而那些錄製假人挑戰影片的人們並不知道他們在輸入景深資訊。

If the researchers hadn't talked about their training system, it would feel like they'd done it all on their own.

如果研究人員們沒有談論過他們的訓練系統，感覺起來就好像完全是他們自己做的。

The remarkable thing about AI systems is that even though they are built on a foundation of human intelligence,

人工智慧最神奇的地方在於，就算它們是建立在人類智慧的基礎上，

they regularly transcend that and do something that surprises us or goes beyond what we thought was possible.

他們經常會勝過基礎，做出讓我們驚豔或是超越我們想像的事物。

One fantastic example of this is the AlphaGo program, which was designed by DeepMind, which is Google's AI lab here in London.

Google 倫敦的 DeepMind 人工智慧實驗室所設計的 AlphaGo 計畫就是個奇妙的例子。

And in 2016 and 2017, it played and beat the human champions of the ancient board game, Go.

在 2016 和 2017 年，它打敗了古代桌遊圍棋的真人冠軍。

There's one particularly famous moment⏤it's now known simply as move 37.

最著名的時刻被稱做第 37 步棋。

It was a move that was so unusual, so counter to human expectations, that the match's commentators thought it was a mistake.

這步棋是如此反常、與人類所預期的如此相反，以至於賽評們認為它是個失誤。

But it wasn't.

但那不是個失誤。

It was a beautiful play that completely undermined Lee's match, and led to AlphaGo winning the game.

那是傑出的一手，完全大勝了 Lee，並讓 AlphaGo 贏得了比賽。

And it was something that humans couldn't teach; it was something that the machine had learned by itself.

那是人類所無法教導的。是機器自己學習而來的。

Yes, it started from a foundation of human intelligence, but it went beyond that.

沒錯，它始於人類智慧的基礎，但它超越了那個基礎。

This, I think, is where people get so excited by AI.

我想這就是人們對人工智慧感到興奮的地方。

We're a long, long way away from building computers that are as flexibly intelligent and, sort of, sophisticated as humans,

雖然我們距離打造出如人類一般成熟、擁有彈性智慧的電腦還有很長一段距離，

but we can still build algorithms and systems that exceed human intelligence, even in very specific domains.

但我們還是能建造出甚至在特定專業領域中超越人類智慧的演算法及系統。

But that's AI at its best.

不過這是人工智慧最好的一面。

The flip side is when an app needs a description of what's in a photo, and the photo-recognizing algorithm just doesn't work.

另外一面就是，當一個應用程式需要描述照片裡有什麼時，照片辨識演算法卻無法運作。

So you get a human being to fill it in, usually through a post on Mechanical Turk.

所以就透過 Mechanical Turk 上的貼文找一個人類來代勞。

That's a very old trick, going all the way back to the machine that gave the site its name.

那是個非常古老的伎倆，追溯至網站依其命名的機器。

The original Mechanical Turk was this guy, a master chess-playing robot, hundreds of years before there was anything we would think of as a computer.

原始的機械土耳其人是他，一個西洋棋機器人大師。在電腦出現的好幾百年前，就已存在。

The Turk could beat most chess players, playing so well that people thought it was a technological marvel.

這個機器人能夠打敗大部分的西洋棋玩家，他下棋下得太好了以至於人們認為他是個科技奇蹟。

But, really, it was just a trick.

但其實他只是個小把戲。

There was a human being inside, hiding under the table and directing the moves from below.

有個人類躲在桌子底下控制著每步棋。

It was a human being dressed up as a machine, a trick no one had thought of until then.

他只是個裝扮成機器人的人類，是個直到當時沒人想出來的把戲。

And as Amazon can tell you, the trick still works.

而就像在 Amazon 看到的，這個把戲還相當有用。

Thanks for watching; I hope you liked the video.

感謝收看，希望你喜歡這部影片。

If you wanna know more about AI, we did a whole video about, sort of, what these changes look like at a social scale, whether AI's destroying jobs or gonna make everything free.

如果你想更深入了解人工智慧，我們做了個關於社會變動的影片，探討了是否人工智慧會毀掉不同工作或使每樣東西免費化。

So, you can check that out here or like and subscribe.

你可以在這裡觀看，或是按喜歡並訂閱。