Placeholder Image

字幕列表 影片播放

  • Every day, a large portion of the population

    每一天,一大群的人們

  • is at the mercy of a rising technology,

    享受著科技帶來的便利

  • yet few actually understand what it is.

    但其實很少人知道這背後到底是哪些科技在支撐著他們的生活

  • Artificial intelligence. You know, HAL 9000

    人工智慧。你知道的,就是 HAL 9000

  • and Marvin the Paranoid Android?

    以及疑心病很重的那個人型機器人 Marvin?

  • Thanks to books and movies,

    多虧了每個世代的書和電影

  • each generation has formed

    各自想像描述了屬於他們

  • its own fantasy of a world ruled

    自己由機器人所統治的

  • -- or at leased served -- by robots.

    或者說至少是由機器人提供服務的世界

  • We've been conditioned to expect flying cars

    我們已經被制約去期待一個飛行汽車出現的

  • that steer clear of traffic

    一個不會塞車的世界

  • and robotic maids whipping up our weekday dinner.

    以及機器人侍者在我們週間用晚餐時清潔打掃

  • But if the age of AI is here,

    但是如果 AI 時代到來時

  • why don't our lives look more like the Jetson's?

    為什麼看起來不會更像是 Jetson 過的生活那個樣貌?

  • Well, for starters, that's a cartoon.

    這個麻,對於初次接觸的人來說,是一部卡通的概念

  • And really, if you've ever browsed Netflix movie suggestions

    而實際上,你若有瀏覽過 Netflix 電影建議清單

  • or told Alexa to order a pizza,

    或者曾經叫過 Alexz 去幫你訂披薩

  • you're probably interacting with artificial intelligence

    你與人工智慧互動的程度

  • more than you realize.

    大概比你認知到的還要多更多

  • And that's kind of the point.

    而這就是重點所在

  • AI is designed so you don't realize

    AI 本來就是應該設計得讓你無法察覺

  • there's a computer calling the shots.

    背後其實是有電腦在決定這個動作的

  • But that also makes understanding what AI is,

    但是也讓我們能夠理解到底什麼是 AI

  • and what it's not, a little complicated.

    什麼又不是,這其中其實有點複雜

  • In basic terms, AI is a broad area of computer science

    但基本上來說,AI 就是一個非常大的電腦科學領域

  • that makes machines seem

    能夠讓機器看起來

  • like they have human intelligence.

    好像擁有人類的智力

  • So it's not only programming a computer to drive a car

    所以這不僅僅是寫一個程式讓電腦能夠遵守交通號誌來駕駛汽車

  • by obeying traffic signals, but it's when that program

    也是一個能夠

  • also learns to exhibit signs of human-like road rage.

    學習模仿人類在駕車時產生的「路怒」的能力

  • As intimidating as it may seem,

    看起來好像很嚇人

  • this technology isn't new.

    但這個並不是新科技了

  • Actually, for the past half-a-century,

    事實上,大概半個世紀前就有了

  • it's been an idea ahead of its time.

    所以這個是相當前衛的一個點子

  • The term "artificial intelligence" was first coined back in

    「人工智慧」一詞在1956年時

  • 1956 by Dartmouth professor John McCarthy.

    由 Dartmouth 的教授 John McCarthy 所創造

  • He called together a group of computer scientists and mathematicians

    他召集了一群電腦和數學科學家

  • to see if machines could learn like a young child does,

    來測試看看電腦是否能夠擁有像是小孩子的學習能力

  • using trial and error to develop formal reasoning.

    藉由嘗試錯誤來發展一個合乎正常的推理能力

  • The project proposal says they'll figure out how to make machines

    這個案子的企劃寫說他們會探究出如何讓機器:「具有語言能力、

  • "use language, form abstractions and concepts,

    具備抽象及概念性思考

  • solve kinds of problems now reserved for humans,

    以及現今人類面臨的眾多不同的問題

  • and improve themselves."

    並且機器能夠自我改進。」等等能力

  • That was more than 60 years ago.

    這已經是超過 60 年之前了

  • Since then, AI has remained for the most part

    至從那時起,AI 就在眾多的大學課堂中及超級機密實驗室中

  • in university classrooms and super secret labs ...

    被廣泛的研究

  • But that's changing.

    但這已經改變了

  • Like all exponential curves, it's hard to tell when a line

    就跟所有的指數函數曲線,是很難判斷一條數學關係曲線

  • that's slowly ticking upwards is going to skyrocket.

    從起初慢慢遞增的狀態,突然竄升的時機點在什麼時候

  • But during the past few years, a couple of factors

    但在過去幾年當中,其中的一些因素

  • have led to AI becoming the next "big" thing:

    導致 AI 成為下一個新顯學

  • First, huge amounts of data are being

    首先,每分鐘被創造出來的數據

  • created every minute. In fact, 90% of the world's data

    急速增加。事實上,世界上百分之九十的資料

  • has been generated in the past two years.

    是過去兩年以內產生的

  • And now thanks to advances in processing speeds,

    而現在多虧了在運算處理能力的進步

  • computers can actually make sense

    電腦能夠更睏素且確實地理解

  • of all this information more quickly.

    所有接收到的資訊

  • Because of this, tech giants and venture capitalists

    因為這樣,科技巨人及創投資本家

  • have bought into AI and are infusing the market

    都進入了 AI 市場並且將資金和新的應用融合

  • with cash and new applications.

    到市場當中

  • Very soon, AI will become a little less artificial,

    很快地,AI 將會變成比較不「人工」

  • and a lot more intelligent.

    而且會更加「智慧」

  • Now the question is: Should you brace yourself for yet

    現在問題變成是:你應該該要接受

  • another Terminator movie, live on your city streets?

    你所生活的城市街上上演電影「魔鬼終結者」嗎?

  • Not exactly. In fact, stop thinking of robots.

    不完全是。事實上,不要朝著機器人的方面去思考

  • When it comes to AI, a robot is nothing more than

    當我們談到 AI 時,機器人往往只是

  • the shell concealing what's actually used

    真正推動這項科技的

  • to power the technology.

    外殼罷了

  • That means AI can manifest itself in many

    也就是說 AI 能夠把他自己實現在

  • different ways. Let's break down the options.

    不同的地方。我們來解析看看有哪些選項

  • First, you have your bots. They're text-based and

    第一,你會有屬於自己的機器人。他們是以文字輸入作為基礎並且

  • incredibly powerful, but they have limitations.

    功能不可思議的強大,但是還是有極限

  • Ask a weather bot for the forecast, and it will tell you

    詢問一個天氣機器人幫你做天氣預報,它會告訴你

  • it's partly cloudy with a high of 57.

    今天天氣是局部多雲,高溫華氏 57 度

  • But ask that same bot what time it is in Tokyo,

    但問同一個機器人現在東京的時間是多少

  • and it'll get a little confused.

    它可能會有點混淆不清你的意思是什麼

  • That's because the bot's creator only programmed it to

    這是因為這個機器人的創造者只將他設定成

  • give you the weather by pulling from a specific data source.

    有能力去存取特定的資料來源以便能告訴你天氣的狀況而已

  • Natural language processing makes these bots

    自然語言處理讓這些機器人

  • a bit more sophisticated.

    變得更有智慧一些

  • When you ask Siri or Cortana

    當你問 Siri 或者 Cortana

  • where the closest gas station is,

    最近的加油站在哪的時後

  • it's really just translating your voice into text,

    它們真的就只是將你的聲音化為文字

  • feeding it to a search engine,

    輸入搜尋引擎而已

  • and reading the answer back in human syntax.

    並且把答案讀取出來用人類的語氣回答你

  • So in other words, you don't have to speak in code.

    所以換句話說,你不用說出「程式語言碼」就可以達到跟電腦溝通的目的

  • At the far end of the spectrum is machine learning,

    在 AI 能力的功能的光譜上的最終端就是機器學習

  • and honestly, it's one of the most exciting areas of AI.

    而說實話,這是 AI 讓人最興奮的領域之一

  • Like a human, a machine retains information

    像一個人類一般,一個機器獲取訊息

  • and becomes smarter over time.

    並且隨時間變得更加聰明

  • But unlike a human, it's not susceptible to things like

    但是跟人不同的地方,它不是非常容易發生像是

  • short-term memory loss, information overload,

    短期記憶的流失、資訊量過載

  • sleep deprivation, and distractions.

    睡眠剝奪以及受到分心

  • But how do these machines actually learn?

    但是這些機器到底是如何進行學習的?

  • Well, while it may be easy for a human to know

    這個麻,人們很容易能夠知道

  • the difference between a cat and a dog,

    一隻狗和一隻貓的不同點

  • for a computer, not so much.

    但對一部電腦來說,不是這麼容易的

  • You see, when you're only considering

    你看,當你只考慮到

  • physical appearance, the difference between

    形體的外表,這些在貓和狗之間

  • cats and dogs can be a little gray.

    的差異就會有點模糊

  • You can say cats have pointed ears

    你可以說一隻貓有尖尖的耳朵

  • and dogs have floppy ears,

    而狗狗有軟趴趴耳朵

  • but those rules aren't universal.

    但是這些規則並非普遍一致適用

  • Between tail length, fur texture, and color,

    在以八的長度、毛的質地,和顏色

  • there are a lot of options,

    有非常多的選擇

  • and that means a lot of tedious rules someone would

    這代表很多生澀難解的規則

  • have to program manually to help a computer

    會需要被寫進電腦程式中幫助電腦

  • spot the difference.

    去判斷它們這之間的差異

  • But remember -- machine learning is about making

    但切記,機器學習是關於

  • machines learn like humans. And like any toddler,

    如何讓機器像人一樣去學習。而像每個寶寶一樣

  • that means they have to learn by experience.

    代表它們需要透過經驗來學習

  • With machine learning, programs analyze

    關於機器學習,電腦程式會分析

  • thousands of examples to build an algorithm.

    上千筆的案例來建立一組演算法

  • It then tweaks the algorithm

    然後它會根據它是否達成目標

  • based on if it achieves its goal.

    來調整演算法

  • Over time, the program actually gets smarter.

    隨著時間,程式的確會變得更聰明

  • That's how machines like IBM's Watson can

    這是像 IBM 的華生電腦一樣的機器如何能夠

  • diagnose cancer, compose classical symphonies,

    診斷癌症、譜出古典樂曲

  • or crush Ken Jennings at Jeopardy.

    或者在 Jeopardy 上擊潰 Ken Jennings 的原因

  • Some programs even mimic the way

    某些程式甚至會模仿人類

  • the human brain is structured,

    大腦的組成方式

  • complete with neural networks that help humans --

    以神經網路姿態出現來幫助人類

  • and now machines -- solve problems.

    以及現今的機器來解決問題

  • Generations have long imagined the ramifications of AI,

    幾個世代以來人類長期一直想像 AI 可能的各種發展

  • visualizing a society where machines seek revenge

    描繪一個機器對人類復仇的社會狀況

  • and wreak havoc on human society.

    導致人類社會的毀滅

  • However, the more logical and pressing question is:

    然而,更符合邏輯的問題將是:

  • How will AI affect your job?

    AI 對你的工作會產生什麼影響?

  • Will it make your work obsolete?

    它將會讓你的工作被取代掉嗎?

  • Just like the Industrial Revolution,

    就如工業革命一樣

  • it's not human versus machine.

    這並不是人類對上機器

  • It's human and machine versus problem.

    而是人類和機器一同面對問題

  • The point is that artificial intelligence

    重點在人工智慧

  • helps you accomplish more in less time,

    會幫助你以更少的時間完成更多的事情

  • taking on the repetitive tasks of your job

    把你工作中重複性的部分分擔掉

  • while you master the strategy and relationships.

    而讓你精於處理事情的策略和人際關係的經營

  • That way, humans can do what they do bestbe human.

    如此一來,人類便可以去做他們最擅長的事情...當一個真正的人類

Every day, a large portion of the population

每一天,一大群的人們

字幕與單字

單字即點即查 點擊單字可以查詢單字解釋