Placeholder Image

字幕列表 影片播放

  • Let me tell you a story.

    首先,我想與大家分享一個故事。

  • It goes back 200 million years.

    時鐘撥回到兩億年前,

  • It's a story of the neocortex,

    我們的故事,

  • which means "new rind."

    與新皮層(neocortex)有關。

  • So in these early mammals,

    早期哺乳動物

  • because only mammals have a neocortex,

    (實際上只有哺乳動物才有新皮層)

  • rodent-like creatures.

    比如齧齒類動物,

  • It was the size of a postage stamp and just as thin,

    擁有一種尺寸和厚度與郵票相當的新皮層,

  • and was a thin covering around

    它像一層薄膜,

  • their walnut-sized brain,

    包覆著這些動物核桃大小的大腦。

  • but it was capable of a new type of thinking.

    新皮層的功能不可小覷, 它賦予動物新的思考能力。

  • Rather than the fixed behaviors

    不像非哺乳類動物,

  • that non-mammalian animals have,

    牠們的行為基本上固定不變,

  • it could invent new behaviors.

    擁有新皮層的哺乳動物能發明新的行為。

  • So a mouse is escaping a predator,

    比如,老鼠逃避天敵的追捕時,

  • its path is blocked,

    一旦發現此路不通,

  • it'll try to invent a new solution.

    牠會嘗試去找新的出路。

  • That may work, it may not,

    最終可能逃之夭夭,也可能落入貓口,

  • but if it does, it will remember that

    但僥倖成功時,牠會記取成功的經驗,

  • and have a new behavior,

    最終形成一種新的行為。

  • and that can actually spread virally

    值得一提的是,這種新近習得的行為,

  • through the rest of the community.

    會迅速傳遍整個鼠群。

  • Another mouse watching this could say,

    我們可以想像,一旁觀望的老鼠會說:

  • "Hey, that was pretty clever, going around that rock,"

    “哇,真是急中生智,居然想到繞開石頭來逃生!”

  • and it could adopt a new behavior as well.

    然後,輕而易舉也掌握了這種技能。

  • Non-mammalian animals

    但是,非哺乳動物

  • couldn't do any of those things.

    對此完全無能為力,

  • They had fixed behaviors.

    牠們的行為一成不變。

  • Now they could learn a new behavior

    準確地說,牠們也能習得新的行為,

  • but not in the course of one lifetime.

    但不是在一朝一夕之間,

  • In the course of maybe a thousand lifetimes,

    可能需要歷經一千個世代,

  • it could evolve a new fixed behavior.

    整個種群才能形成一種新的固定行為。

  • That was perfectly okay 200 million years ago.

    在兩億年前的蠻荒世界, 這種進化節奏並無大礙。

  • The environment changed very slowly.

    那時,環境變遷步履蹣跚,

  • It could take 10,000 years for there to be

    大約每一萬年,

  • a significant environmental change,

    才發生一回滄海桑田的巨變,

  • and during that period of time

    在這樣一個漫長的時間跨度裏,

  • it would evolve a new behavior.

    動物才形成了一種新的行為。

  • Now that went along fine,

    往後,一切安好。

  • but then something happened.

    直到,禍從天降。

  • Sixty-five million years ago,

    時間快進到6500萬年前,

  • there was a sudden, violent change to the environment.

    地球遭遇一場突如其來的環境遽變,

  • We call it the Cretaceous extinction event.

    後人稱之為“白堊紀物種大滅絕”。

  • That's when the dinosaurs went extinct,

    恐龍遭受滅頂之災;

  • that's when 75 percent of the

    75%的地球物種

  • animal and plant species went extinct,

    走向滅絕;

  • and that's when mammals

    而哺乳動物

  • overtook their ecological niche,

    趁機佔領了其他物種的生存地盤。

  • and to anthropomorphize, biological evolution said,

    我們可以假託這些哺乳動物的口吻, 來評論這一進化過程:

  • "Hmm, this neocortex is pretty good stuff,"

    “唔,關鍵時候我們的新皮層真派上用場了。”

  • and it began to grow it.

    此後,新皮層繼續發育。

  • And mammals got bigger,

    哺乳動物個頭也日漸見長,

  • their brains got bigger at an even faster pace,

    大腦容量迅速擴大,

  • and the neocortex got bigger even faster than that

    其中新皮層的發育堪稱突飛猛進,

  • and developed these distinctive ridges and folds

    已經逐步形成獨特的溝回和褶皺,

  • basically to increase its surface area.

    這可以進一步增加其表面積。

  • If you took the human neocortex

    人類的新皮層,

  • and stretched it out,

    如果充分展開平鋪,

  • it's about the size of a table napkin,

    尺寸可達一張餐巾大小。

  • and it's still a thin structure.

    但它仍然保持了纖薄的結構,

  • It's about the thickness of a table napkin.

    厚度也與餐巾不相上下。

  • But it has so many convolutions and ridges

    外形曲折複雜,呈現千溝萬壑,

  • it's now 80 percent of our brain,

    新皮層已佔據大腦體積的80%左右,

  • and that's where we do our thinking,

    不僅肩負思考的重任,

  • and it's the great sublimator.

    還約束和昇華個人的行為。

  • We still have that old brain

    今天,我們的大腦

  • that provides our basic drives and motivations,

    仍然製造原始的需求和動機。

  • but I may have a drive for conquest,

    但是,對於我們內心狂野的征服欲望,

  • and that'll be sublimated by the neocortex

    這個新皮層起著春風化雨、潤物無聲的作用,

  • into writing a poem or inventing an app

    最終將這種欲望化作創造詩歌、開發APP、

  • or giving a TED Talk,

    甚至是發表TED演講這樣的文明行為。

  • and it's really the neocortex that's where

    對於這一切,

  • the action is.

    新皮層功不可沒。

  • Fifty years ago, I wrote a paper

    50年前,我完成了一篇論文,

  • describing how I thought the brain worked,

    探究大腦的工作原理,

  • and I described it as a series of modules.

    我認為大腦是一系列模塊的有機結合。

  • Each module could do things with a pattern.

    每個模塊按照某種模式各司其職,

  • It could learn a pattern. It could remember a pattern.

    但也可以學習、記憶新的模式,

  • It could implement a pattern.

    並將模式付諸應用。

  • And these modules were organized in hierarchies,

    這些模式以層級結構進行組織,

  • and we created that hierarchy with our own thinking.

    當然,我們借助自己的思考 假設了這種層級結構。

  • And there was actually very little to go on

    50年前,由於各種條件限制,

  • 50 years ago.

    研究進展緩慢,

  • It led me to meet President Johnson.

    但這項成果使我獲得了 約翰遜總統的接見。

  • I've been thinking about this for 50 years,

    50年來,我一直潛心研究這個領域,

  • and a year and a half ago I came out with the book

    就在一年半前,我又發表了一部新的著作

  • "How To Create A Mind,"

    ——《心智的構建》。

  • which has the same thesis,

    該專著探討了同一個課題,

  • but now there's a plethora of evidence.

    幸運的是,我現在擁有充足的證據支撐。

  • The amount of data we're getting about the brain

    神經科學為我們貢獻 大量有關大腦的數據,

  • from neuroscience is doubling every year.

    還在以逐年翻倍的速度劇增;

  • Spatial resolution of brainscanning of all types

    各種腦部掃描技術的空間解析度,

  • is doubling every year.

    也在逐年翻倍。

  • We can now see inside a living brain

    現在,我們能親眼窺見活體大腦的內部,

  • and see individual interneural connections

    觀察單個神經間的連接,

  • connecting in real time, firing in real time.

    目睹神經連接、觸發的實時發生。

  • We can see your brain create your thoughts.

    我們親眼看到大腦如何創造思維,

  • We can see your thoughts create your brain,

    或者反過來說,思維如何增強和促進大腦,

  • which is really key to how it works.

    思維本身對大腦進化至關重要。

  • So let me describe briefly how it works.

    接下來,我想簡單介紹大腦的工作方式。

  • I've actually counted these modules.

    實際上,我統計過這些模塊的數量。

  • We have about 300 million of them,

    我們總共有大約三億模塊,

  • and we create them in these hierarchies.

    分佈在不同的層級中。

  • I'll give you a simple example.

    讓我們來看一個簡單的例子。

  • I've got a bunch of modules

    假設我有一組模塊,

  • that can recognize the crossbar to a capital A,

    可以識別大寫字母“A”中間的短橫線,

  • and that's all they care about.

    它們的主要職責就在於此。

  • A beautiful song can play,

    無論周遭播放著美妙的音樂,

  • a pretty girl could walk by,

    還是一位妙齡女郎翩然而至,

  • they don't care, but they see a crossbar to a capital A,

    它們都渾然不覺。但是,一旦發現“A”的短橫線,

  • they get very excited and they say "crossbar,"

    它們就興奮異常,異口同聲喊出:“短橫線!”

  • and they put out a high probability

    同時,它們立即報告神經軸突,

  • on their output axon.

    識別任務已經順利完成。

  • That goes to the next level,

    接下來,更高級別的模塊——

  • and these layers are organized in conceptual levels.

    概念級別的模塊,將依次登場。

  • Each is more abstract than the next one,

    級別越高,思考的抽象程度越高。

  • so the next one might say "capital A."

    例如,較低的級別可識別字母“A”,

  • That goes up to a higher level that might say "Apple."

    逐級上升後,某個級別能識別“APPLE”這個單詞。

  • Information flows down also.

    同時,信息也在持續傳遞。

  • If the apple recognizer has seen A-P-P-L,

    負責識別“APPLE”的級別,發現A-P-P-L時,

  • it'll think to itself, "Hmm, I think an E is probably likely,"

    它會想:“唔,我猜下一個字母應該是E吧。”

  • and it'll send a signal down to all the E recognizers

    然後,它會將信號傳達到 負責識別“E”的那些模塊,

  • saying, "Be on the lookout for an E,

    並發出預警:“嘿,各位注意,

  • I think one might be coming."

    字母E就要出現了!”

  • The E recognizers will lower their threshold

    字母“E”的識別模塊於是降低了閥值,

  • and they see some sloppy thing, could be an E.

    一旦發現疑似字母,便認為是“E”。

  • Ordinarily you wouldn't think so,

    當然,這並非通常情況下的處理機制,

  • but we're expecting an E, it's good enough,

    但現在我們正在等待“E”的出現, 而疑似字母與它足夠相似,

  • and yeah, I've seen an E, and then apple says,

    所以,我們斷定它就是“E”。

  • "Yeah, I've seen an Apple."

    “E”識別後,“APPLE”識別成功。

  • Go up another five levels,

    如果我們再躍升五個級別,

  • and you're now at a pretty high level

    那麼,在整個層級結構上,

  • of this hierarchy,

    就到達了較高水平。

  • and stretch down into the different senses,

    這個水平上,我們具有各種感知功能,

  • and you may have a module that sees a certain fabric,

    某些模塊能夠感知特定的布料質地,

  • hears a certain voice quality, smells a certain perfume,

    辨識特定的音色,甚至嗅到特定的香水味,

  • and will say, "My wife has entered the room."

    然後告诉我:妻子剛進到房间!

  • Go up another 10 levels, and now you're at

    再上升10級,

  • a very high level.

    我們就到達了一個很高的水平,

  • You're probably in the frontal cortex,

    可能來到了額葉皮層。

  • and you'll have modules that say, "That was ironic.

    在這兒,我們的模塊已經能夠臧否人物了,

  • That's funny. She's pretty."

    比如:這事有點滑稽可笑!她真是秀色可餐!

  • You might think that those are more sophisticated,

    大家可能覺得,這整個過程有點複雜。

  • but actually what's more complicated

    實際上,更讓人費解的是

  • is the hierarchy beneath them.

    是這些過程的層級結構。

  • There was a 16-year-old girl, she had brain surgery,

    曾經有位16歲的姑娘,當時正接受腦部手術。

  • and she was conscious because the surgeons

    由於手術過程中醫生需要跟她講話,

  • wanted to talk to her.

    所以就讓她保持清醒。

  • You can do that because there's no pain receptors

    保持清醒的意識,這對於手術並無妨礙,

  • in the brain.

    因為大腦內沒有痛覺感受器。

  • And whenever they stimulated particular,

    我們驚奇地發現,當醫生刺激新皮層上

  • very small points on her neocortex,

    某些細小區域時,就是圖中的紅色部位,

  • shown here in red, she would laugh.

    這個姑娘就會放聲大笑。

  • So at first they thought they were triggering

    起初,大家以為,

  • some kind of laugh reflex,

    可能是因為觸發了笑反應神經。

  • but no, they quickly realized they had found

    他們很快意識到事實並非如此,

  • the points in her neocortex that detect humor,

    這些新皮層上的特定區域能夠理會幽默,

  • and she just found everything hilarious

    只要醫生刺激這些區域,

  • whenever they stimulated these points.

    她就會覺得所有的一切都滑稽有趣。

  • "You guys are so funny just standing around,"

    “你們這幫人光站在那裏,就讓人想笑。”

  • was the typical comment,

    那位姑娘典型的解釋道。

  • and they weren't funny,

    我們知道,這個場景並不滑稽可笑,

  • not while doing surgery.

    因為大家都在進行緊張的手術。

  • So how are we doing today?

    現在,我們又有哪些新的進展呢?

  • Well, computers are actually beginning to master

    計算機日益智能化,

  • human language with techniques

    利用功能類似新皮層的先進技術,

  • that are similar to the neocortex.

    它們可以學習和掌握人類的語言。

  • I actually described the algorithm,

    我曾描述過一種算法,

  • which is similar to something called

    與層級隱含式馬爾可夫模型類似,

  • a hierarchical hidden Markov model,

    (馬爾可夫模型是用於自然語言處理的統計模型)

  • something I've worked on since the '90s.

    上世紀90年以來我一直研究這種算法。

  • "Jeopardy" is a very broad natural language game,

    “Jeopardy”(危境)是一個 自然語言類的智力競賽節目,

  • and Watson got a higher score

    IBM研發的沃森計算機在比賽中

  • than the best two players combined.

    勇奪高分,總分超過兩名最佳選手的總和。

  • It got this query correct:

    連這個難題都被它輕鬆化解了:

  • "A long, tiresome speech

    “定義:由起泡的派餡料發表的冗長而乏味的演講。

  • delivered by a frothy pie topping,"

    請問:這定義的是什麼?”

  • and it quickly responded, "What is a meringue harangue?"

    它迅速回答道:愛開腔的蛋白霜。

  • And Jennings and the other guy didn't get that.

    而詹尼斯和另外一名選手卻一頭霧水。

  • It's a pretty sophisticated example of

    這個問題難度很大,極富挑戰性,

  • computers actually understanding human language,

    向我們展示了計算機 正在掌握人類的語言。

  • and it actually got its knowledge by reading

    實際上,沃森是通過廣泛閱讀維基百科

  • Wikipedia and several other encyclopedias.

    及其他百科全書來發展語言能力的。

  • Five to 10 years from now,

    5至10年以後,

  • search engines will actually be based on

    我們的搜索引擎

  • not just looking for combinations of words and links

    不再只是搜索詞語和鏈接這樣的簡單組合,

  • but actually understanding,

    它會嘗試去理解信息,

  • reading for understanding the billions of pages

    通過涉獵浩如煙海的互聯網和書籍,

  • on the web and in books.

    攫取和提煉知識。

  • So you'll be walking along, and Google will pop up

    想像有一天,你正在悠閒地散步,

  • and say, "You know, Mary, you expressed concern

    智能設備端的 Google 助理突然和你說:

  • to me a month ago that your glutathione supplement

    “瑪麗,你上月提到,正在服用的谷胱甘肽補充劑

  • wasn't getting past the blood-brain barrier.

    因為無法透過血腦屏障,所以暫時不起作用。

  • Well, new research just came out 13 seconds ago

    告訴你一個好消息!就在13秒鐘前,

  • that shows a whole new approach to that

    一項新的研究成果表明,

  • and a new way to take glutathione.

    可以透過一个新的途徑來補充谷胱甘肽。

  • Let me summarize it for you."

    讓我給你概括一下這個報告。”

  • Twenty years from now, we'll have nanobots,

    20年以後,我們將迎來奈米機器人,

  • because another exponential trend

    目前,科技產品正在日益微型化,

  • is the shrinking of technology.

    這一趨勢愈演愈烈。

  • They'll go into our brain

    科技設備將通過毛細血管

  • through the capillaries

    進入我們的大腦,

  • and basically connect our neocortex

    最終,將我們自身的新皮層

  • to a synthetic neocortex in the cloud

    與雲端的人工合成新皮層相連,

  • providing an extension of our neocortex.

    使它成為新皮層的延伸和擴展。

  • Now today, I mean,

    今天,

  • you have a computer in your phone,

    智慧型手機都內置了一台計算機。

  • but if you need 10,000 computers for a few seconds

    假如我們需要一萬台計算機,

  • to do a complex search,

    在幾秒鐘內完成一次複雜的搜索,

  • you can access that for a second or two in the cloud.

    我們可以通過訪問雲端來獲得這種能力。

  • In the 2030s, if you need some extra neocortex,

    到了2030年,當你需要更加強大的新皮層時,

  • you'll be able to connect to that in the cloud

    你可以直接從你的大腦連接到雲端,

  • directly from your brain.

    來獲得超凡的能力。

  • So I'm walking along and I say,

    舉個例子,我正在漫步,遠遠看到一個人。

  • "Oh, there's Chris Anderson.

    “老天,那不是克里斯.安德森(TED主持人)嗎?

  • He's coming my way.

    他正朝我這邊走來。

  • I'd better think of something clever to say.

    我要抓住這個機遇,一鳴驚人!

  • I've got three seconds.

    但是,我只有三秒鐘,

  • My 300 million modules in my neocortex

    我新皮層的三億個模塊

  • isn't going to cut it.

    顯然不夠用。

  • I need a billion more."

    我需要借來10億模塊增援!”

  • I'll be able to access that in the cloud.

    於是,我會立即連通雲端。

  • And our thinking, then, will be a hybrid

    我的思考,綜合了生物體和非生物體

  • of biological and non-biological thinking,

    這兩者的優勢。

  • but the non-biological portion

    非生物部分的思考能力,

  • is subject to my law of accelerating returns.

    將受益於“加速回報定律”,

  • It will grow exponentially.

    這是說,科技帶來的回報 呈指數級增長,而非線性。

  • And remember what happens

    大家是否還記得,上次新皮層大幅擴張時

  • the last time we expanded our neocortex?

    發生了哪些重大變化?

  • That was two million years ago

    那是200萬年前,

  • when we became humanoids

    我們那時還只是猿人,

  • and developed these large foreheads.

    開始發育出碩大的前額。

  • Other primates have a slanted brow.

    而其他靈長類動物的前額向後傾斜,

  • They don't have the frontal cortex.

    因為牠們沒有額葉皮層。

  • But the frontal cortex is not really qualitatively different.

    但是,額葉皮層並不意味著質的變化;

  • It's a quantitative expansion of neocortex,

    而是新皮層量的提升,

  • but that additional quantity of thinking

    帶來了額外的思考能力,

  • was the enabling factor for us to take

    最終促成了質的飛躍。

  • a qualitative leap and invent language

    我們因而能夠發明語言,

  • and art and science and technology

    創造藝術,發展科技,

  • and TED conferences.

    並舉辦TED演講,

  • No other species has done that.

    這都是其他物種難以完成的創舉。

  • And so, over the next few decades,

    我相信未來數十年,

  • we're going to do it again.

    我們將再次創造偉大的奇蹟。

  • We're going to again expand our neocortex,

    我們將借助科技,再次擴張新皮層,

  • only this time we won't be limited

    不同之處在於,

  • by a fixed architecture of enclosure.

    我們將不再受到頭顱空間的局限,

  • It'll be expanded without limit.

    意味著擴張並無止境。

  • That additional quantity will again

    隨之而來的量的增加

  • be the enabling factor for another qualitative leap

    在人文和科技領域,

  • in culture and technology.

    將再次引發一輪質的飛躍。

  • Thank you very much.

    謝謝大家!

  • (Applause)

    (掌聲)