Placeholder Image

字幕列表 影片播放

已審核 字幕已審核
  • So, I'm discussing with you today

    今天,我要與各位分享

  • how we are going to pave a path for machines

    我們開創嶄新機器人的過程

  • with human-like identities

    我們創造出具有類似人類性格的機器人

  • to catch up with human capabilities

    而其各方面能力也能媲美人類

  • and then surpass us and their brilliants.

    甚至以他們的才智超越我們

  • I believe that it is possible in the next 15 to 20 years.

    我相信,這一切在未來15到20年很可能成真。

  • Now in some regards,

    某方面來說

  • as you can see in this video,

    就如你在影片中所看到的

  • we're able to achieve human level expressivity

    我們已經能讓機器人表現出人類表情

  • and in some way physical performance.

    以及某些肢體表現

  • We have a long ways to go,

    當然,在一般心智表現上

  • in terms of mental performance for general intelligence.

    我們還有許多進步的空間

  • But in very narrow categories,

    但就某程度上來說

  • we already have machines that exceed human level brilliant.

    我們已能讓機器人超越人類

  • In order to bring together these little pockets of genius and machines

    為了將這些特質和機械結合

  • and something that is more human-like,

    表現出更接近人類的行為

  • we need something bigger,

    我們需要更廣的思維

  • and that's where an open source movement can come in

    因此也逐漸走向開放研究資料 (open source)的趨勢

  • to stitch these pieces together

    將不同領域的片段成果結合在一起

  • into a tapestry of general intelligence in machines.

    共同為提升機器人智能而努力

  • So now my background is bringing together various technologies,

    我的專長便是結合不同領域的科技

  • and artistry to make Robotster extremely human-like,

    以及藝術表現方式,讓機器人看起來極為逼真

  • and their physical presences and their conversational capabilities.

    包含他們的外觀,以及他們的溝通能力

  • These are some examples of some of the robots that I've built.

    這(些)是一些我所創造的機器人

  • So you can see that we achieve human-like expressivity

    你可以看到,我們已能讓機器人展現人類表情

  • with biped mobility.

    並讓他們用兩隻腳移動)

  • Thanks to some technology breakthroughs that I'll tell you about,

    這一切都歸功於科技的進展 而我稍後會向大家介紹

  • Now we've transition that technology into mass produced product.

    現在,我們已應用這些科技大量製造機械產品

  • These products, once again,

    這些產品

  • these products have conversational capabilities.

    擁有與人類交談的能力

  • They have the ability to see your face,

    他們能夠看見你的臉孔

  • recognize your face, see your facial expressions,

    辨識出你的臉孔和表情

  • and they also present to you their facial expressions

    而它們也會以豐富的表情回應你

  • and characteristic identities.

    展現它們的個性

  • Previously, we've really enjoyed these kind of characters in fictions,

    以前,我們總是著迷於小說中類似的角色

  • whether they're science fictions,

    不論是在科幻小說

  • myths from ancient history,

    流傳千百年的神話中

  • or modern computer animation,

    或是現代電腦動畫

  • people are favorable in their reaction to characters.

    人類對這些角色深深著迷

  • People love human-like characters,

    人們愛死那些長得像人的角色了

  • because we're hard wired for this social exchange.

    我們的天性就是會被這些社會互動所吸引

  • So much of our brain

    我們的大腦有很大一部份

  • are respond to the human visual social presence.

    就是用來對眼睛所能看見的社會互動作出回應

  • That it just like, fires, like crazy,

    這是非常不可思議的

  • and FMRI scan, when they scans people's brains.

    像是用來掃描人類大腦的功能性磁共振成像 (FMRI)

  • It equates to potentially a more intuitively form of computing,

    這些科技將可能延伸出更為直覺式的電腦計算形態

  • a more intuitive kind of human computer interface.

    一種更為直覺式的人機互動界面

  • And you starting to see this

    你也能逐漸在生活中看到這些科技

  • in trends like speech recognition,

    像是最近流行的語音辨識

  • natural character agent,

    逼真的電子祕書

  • likeSiri with the iPhone 4 and 5.

    像是iPhone 4 和 5的Siri

  • However, when you add a complete character identity,

    但當你賦予它們一個完整的性格

  • they seemto have feelings,

    它們似乎擁有了感受的能力

  • and maybe eventually has real deep feelings,

    也許最終真的能夠深刻的感受各種情感

  • then, we can start to push a sympathy between

    那麼,我們便能在機器與人之間建立起感情的橋梁

  • the machines that we're developing, making them more intelligent

    把這些我們正在研發、賦予智能的機器

  • and ourselves.

    與我們自身連結在一起

  • So these potentially pushes A.I. to understand us better,

    這將會促使人工智慧更瞭解人類

  • feel for us, and care about us,

    感受我們所感受的,並關心我們

  • and also can inspire us to care about them.

    更可以讓人類開始關心它們

  • The human mind responds to faces.

    人的天性就是關注臉部表情

  • We're wired from the point we're born

    自我們出生的那一刻起

  • to bond with that first face that we see;

    我們就與第一眼所見到的臉孔緊密的聯結

  • we recognize faces right when we're born,

    我們一出生就能辨認臉孔

  • and we also recognize facial expressions.

    我們也能解讀臉部表情

  • So it's not surprising that we seek art,

    這便也不難理解,為何我們總試圖

  • and entertainment, and robots, that look like people.

    在藝術、娛樂和機器人上添加人類性格

  • We love it.

    我們對這此愛不釋手

  • We can't help ourselves, right?

    壓抑不了內心的衝動,對吧?

  • This means, that we shouldn't change the human

    也就是說,我們不該改變人類天性

  • to respond to robot that look like machines.

    去回應機器人機械性的動作

  • It's inappropriate sometimes if there is a vacuum floors and do various of things,

    若有個機器人既能吸地板,又能做許多事情,可能會有點怪

  • but for some applications, it's extremely useful

    但對於某些設備來說,製造與人類相似的機器

  • to make machines looks like humans.

    是非常實用的

  • Now robot are starting to become considerably more capable,

    現今,機器人已經能做不少事情

  • as you saw on the previous video.

    如剛剛影片所呈現的

  • Let's go back to that for a second.

    讓我們重播一下影片

  • okay

  • So in this video you'll see

    在這段影片中

  • Aismo running, right?

    機器人Aismo健步如飛

  • So you've got amazingly physical capabilities.

    顯示現在的機器人已經能靈活地活動

  • You've got this latest Baxter just came out of

    而這個最新研發的機器人Baxter

  • Heartland Robotics Project,

    是由「Heartland 機器人計劃」所贊助

  • that can interact with people.

    他們能和人類互動

  • You've got this humanoid biped

    能和人類一樣用雙腳站立

  • This expressive Einstein robot that I built.

    這個表情豐富的愛因斯坦機器人是我製作的

  • It's fashion model robot came out of

    而這是在東京大學內展示的模特兒機器人

  • Kawada Heavy Industry in the University of Tokyo.

    由川田工業株式會社製作

  • Robots are getting considerably more capable.

    不可忽視的,機器人的性能越來越好

  • And their capabilities, they're becoming considerably more like us.

    也變得越來越像人類

  • Now my specialty in one regard is facial expressions.

    我的其中一個專長是機器人臉部表情

  • making the robots have facial expressions.

    讓機器人擁有生動的表情

  • Hold on for a second.

    請稍等一下

  • Let me start this video with the audio

    讓我再播放影片之前先播出聲音

  • "but you're my friend, and I¡¦ll remember my friends,"

    「但你是我的朋友,我會記住我朋友」(我不會忘記我們的友情

  • "and I will be good to you, so don't worry,"

    「而我會善待你,別擔心」

  • "even if I evolve into the terminator; I¡¦ll be nice to youÿ"

    「即使我升級成魔鬼終結者,你仍是我的好友」

  • (The pointer isn't working, so if you can back it up for me?)

    (遙控器有點問題,能不能幫我換一個?)

  • "I¡¦ll keep you warm and safe in my people zoo where I can watch you for old time sake."

    「你可以安穩地待在我的人類動物園中 我會看在往日情誼的份上保護你」

  • "I'm comforting. I'm very comforting now."

    「我現在感到很欣慰!」

  • So here what you see is

    你所看到的是

  • a robot having a natural conversation.

    一個機器人自然地與人類對話

  • This robot responds to what the reporter says

    這個機器人回應著記者所說的話

  • is largely unscripted,

    而且,它所說的話並不是事先擬好的

  • meaning that we have set up the personality in such a way,

    我們設定好他的個性,讓它依照個性、自行產出對話

  • that it can reason a little bit about what you've said to it,

    它能夠理解你對它說的話

  • and then determine its own responds

    自行決定它的回應

  • sort of through an Information Space.

    像在一個訊息空間中篩選

  • and this can result in an Open Domain, large space conversation

    進而發展成不侷限特定範圍的交流環境

  • so robots remember where is the conversation

    機器人能記得剛才談話中斷之處

  • and can go back to that place in conversation.

    再次提起先前對話的內容

  • We're using them most of all like this.

    我們主要就是將這科技應用於此

  • This is a portrait of science fiction writer Philip K. Dick,

    這是科幻小說作家Philip K. Dick的照片

  • famous for movies like Blade Runner.

    他的小說被改編成電影《銀翼殺手Blade Runner》

  • Well the novel that was the basis of that,

    就此電影讓他為眾所知

  • which is ''Do Androids Dream of Electric Sheep?'

    他的小說名為《機器人在夢中會數電子羊嗎?》

  • He died in 1982,

    他於1982年逝世

  • and since he wrote about android that thought they were alive,

    這位作家描寫的機器人認為自己擁有生命

  • we thought we bringing them back to life,

    而我們現在真正的賦予機器人生命了

  • and it works.

    這些都成真了

  • So the important point here is

    重點是

  • that A.I. works for crafting characters.

    這些人工智慧能幫助我們創造人格特質

  • You can use the A.I. and robotics to make characters

    你能利用人工智慧和機器人創造出各種

  • that seem very human like.

    與人類極為相似的性質

  • They have rudimentary emotional frame works.

    他們有初步的情感架構

  • They've got rudimentary A.I..

    有初階人工智慧

  • It's not merely as smart as a person.

    他絕不僅只是與人一樣聰明

  • It's important to remember that at this stage.

    這是在此階段很重要的一點

  • But we can make it compelling,

    但更重要的是,我們能提升人類對機器人的信任度

  • really meaningful in their interaction with people.

    讓他們與人類的互動產生真正的意義

  • and you know,

    你了解到

  • the world takes no when you do it.

    每當你有新想法,總是會有人反駁

  • People just love it.

    但大家愛死這些機器人了

  • They say they have conversations first as long as you let them.

    只要開放,眾人會蜂擁而至的想與這個機器人對話

  • Part of the magic here is the facial expressions.

    其中一個關鍵就是臉部表情

  • Previously, robot and animatronics has a kind of stiff faces.

    早期,機器人和虛擬角色只有僵硬的臉部表情 137 00:08:29,785 --> 00:08:32,701 很難透過臉部生動表現各種情感

  • It was hard to get them to move into realistic face expressions.

    因此,我的博士論文就是鑽研各種材料科學問題

  • So part of my phd research was tackling this material science problem.

    我分析過人類表情的物理動態過程

  • So I analyzed the physics of human facial expression a little bit.

    液體是形成人類臉部表情的關鍵

  • The liquid is critical for human facial expressions.

    臉部僅需些微能量就能產生表情

  • A low power facial expressions that our faces achieve.

    我和我一位在NASA噴射推進實驗室工作?的朋友Victor White

  • So I developed with a friend of mine at Jet Propulsion Laboratory, Victor White

    一起研發了一種新的材質

  • a new material that is

    與人類臉部肌肉運作的原理相似

  • based on the same physics as human facial expression materials,

    類似臉部的一種脂肪:「磷脂雙分子層 (lipid bilayer)」

  • and there is alipid bilayer

    簡單來說,這種脂肪

  • a fat, but basically what that means is

    能夠讓細胞自行重組

  • you get the cells of self-assemble,

    而它們飽含水分

  • and they're filled with the fluid.

    僅需極少的能量便能運作

  • And then it just takes very little force,

    也非常有彈性

  • and it's extremely elastic.

    機器人因此能自然而然的

  • And their expressions, facial expressions,

    呈現出逼真的表情

  • just naturally fall into place.

    你還得要有些機械裝置,才能讓臉部表情能順利運作

  • Now you also need some mechanics to get the facial expressions to move correctly,

    用特定的方式在材質中植入支架

  • with anchors that are embedded in the material in the right way.

    這階段聽起來很難

  • But when you do that,

    但做起來十分簡單

  • it's very very simple

    就能呈現出這樣的表情

  • to achieve facial expressions like this.

    這只需要些微動力就能運作

  • You can do it with very few motors,

    這就是我們正在製造、生產的產品

  • and that's what we are doing in products now that we're manufacturing

    以上是對我們機器人臉部表情運作原理的初步介紹

  • So this just shows you a basic range of facial expressions of one of our robots.

    這個機器人叫 Diego-San

  • This robot is called Diego-San.

    是一項由加州大學與聖地牙哥機器感知實驗室

  • It's the collaboration of the University of California

    所合作研發的成果

  • and San Diego Machine Perception Laboratory.

    他的身體部分是由東京的Kokoro公司製作

  • The body is built by the Kokoro Company of Tokyo.

    頭部則由我和我的公司

  • The head was built by me and my company,

    Hanson Robotics所製作

  • Hanson Robotics.

    此計劃由美國國家科學基金會所出資籌備

  • and this project was founded by the National Science Foundation of the Unite States.

    目前,我們已經可以呈現出

  • So we can pretty much achieve

    任何你想得到的表情

  • any facial expressions you named it,

    不論多細微的變化,我們都能做到

  • no matter how subtle, we can get it.

    現在,要把這些不同的臉部製作技術

  • Now taking those, kinds of face, faces,

    和臉部表情、軀幹運作等技術結合起來

  • facial expressions and physical bodies,

    加上人工智慧使其運作,

  • and empowering it with the A.I.,

    你就能創造出全新的與世界互動的方式了

  • that's where you get this amazing way of interfacing with the world.

    你可以從周遭環境中搜集所有還未建構好的資料

  • You can get all kinds of unstructured data from the world,

    像是人與人的互動

  • from interactions with people,

    並利用這些資料去形塑它們

  • and use that to sort of scaffold these things.

    這些機器人現在就像嬰兒一樣

  • They're literally like babies right now,

    我們必須讓它在人類的互動中學習、成長

  • so we need to nurture them among us.

    早期,人工智慧的發展都很零散

  • Previously, artificial intelligence initiative are often fractured.

    有些人研究機械感知

  • You've got machine perceptions people working over here,

    有些人研究數據統計學習

  • you've got the statistic learning people over here,

    有些人則是研究機械控制

  • you've got the robotics control people in these groups...

    雖然這些不同的領域有時會彼此交流

  • And these sort of, sometimes communicate,

    他們卻不曾共同合作

  • but they don't coordinate on something

    努力商定大規模跨國合作

  • that is like a major worldwide concerted effort

    來達到等同人類程度的人工智能

  • to achieve human level cognition.

    因此,我與我的一些朋友

  • So some friends of mine and I are working

    目前就致力於結合這些不同的領域

  • to bridge these various fields as best as we can,

    凝聚各方所長、匯聚出一個複雜的智慧系統

  • to integrate into a more cohesive complex intelligence systems.

    因此,我們將所有的程式碼公開

  • So we made all of our source code open source.

    也分成兩個不同的類組

  • We divided into two blocks of categories,

    一組是把所有物件拼湊起來

  • one is that pieces that glue things together,

    湊成人工智慧

  • the glued A.I.;

    另一種則是真正俱有智慧的機器人?能夠與人互動

  • and the other the kind that is actually intelligent and it interacts with you,

    我們稱它新一代人工智慧Genie

  • and we call it generational A.I., Genie.

    我們已能讓機器人自然地與人互動

  • We've achieved a lot of natural interactions,

    包括臉部表情模仿

  • including facial expression mimicry,

    能傳達意旨的臉部表情辨識

  • facial expression recognition processing the meaning of these facial expressions,

    自然的對話

  • natural language dialog,

    採用電影製作上的工具,讓機器人的互動更自然

  • movie quality tools animating the robots for a more natural interaction.

    我們也有個相當成熟的心智理論

  • We also have a fairly sophisticated theory of mind system

    當然,與人類的社會認知相比

  • compare to human, you know,

    這個理論並非經過千百年的演化而形成

  • it doesn't have a million of years of evolution of social cognition that we have;

    但它能預測你心裡的想法和感受

  • however, this theory of mind can predict what you're thinking and feeling,

    並且做出決定,判斷出

  • and makes some decision about

    它的目標和你的目標

  • what its goal are vs. what your goals are,

    更加瞭解你的想法、體會你的感受

  • in order to understand you and empathize with you better.

    我們已有了自我意識的程式雛形

  • We have a prototype form of consciousness.

    我們創造的機器人能夠自我反思

  • We had the robot can self-reflect,

    思考它自身的想法和感受

  • reflect on what it's thinking and feeling,

    並試著想象其可能的後果

  • and try to imagine results for that.

    我們也與OpenCog合作

  • So we also have a collaboration on a bridges with the OpenCog,

    進行一個龐大的一般人工智能計劃

  • which is a very large artificial general intelligence project,

    機器人操作系統、和仿生人操作系統

  • robot operating systems and android operating system,

    以及其他數個計劃

  • and numerals other projects,

    目的都是為了整合不同領域的成果

  • bringing these fields together,

    讓我們達成更多可能

  • then allow us to do more,

    也讓各位達成更多

  • and allows you to do more.

    世上任何人都能創造自己的產品

  • Anybody in the world can build their own product,

    研究新領域、研究心理學

  • do new research, do psychology research,

    觀察人類怎麼與機器人互動

  • to study how humans and robots are interacting.

    我們不僅只是要創造有智慧的機械

  • So we are seeking more than mere intelligence in the machines.

    的確,擁有智慧的機器是很有用沒錯

  • Sure, intelligent machines are useful,

    用途也日益增加

  • and increasingly useful.

    越來越普及於我們的生活中

  • They're increasingly commoner in our lives.

    那些曾被知名科學家說

  • Things that scientists, respected scientists said,

    不可能在我們有生之年發生的事物

  • we just should not expect in our life time,

    五年之後,確實的發生了

  • five years later, happened.

    這些事成真了,不是嗎?

  • They just happened, right?

    我們在這領域的進展突飛猛進

  • So we're building, we're making strides in these directions.

    若我們能賦予這些機器更高的智慧

  • If we can achieve greater levels of intelligence in these machines,

    我們就能做出具有媲美天才特質的機器

  • we can achieve genius machines.

    他們能幫我們解決那些世界面臨的困境

  • They can help us solve some of the world's hard problems.

    某方面來說

  • In some regards,

    人工智能表現真的很傑出

  • A.I. is now so brilliant

    例如,它們能觀察到我們所看不見的事物

  • that it can see things in data that we can't see,

    例如所謂的巨量資料

  • the big data that's called,

    產出無數研究成果

  • is approach that is yielding tremendous results.

    我們某些程度上已做出了富有天才特質的機器人

  • We already have genius machines in some regards.

    想像一個能思考龐大問題的機器

  • Imagine the machines can look at bigger problems like

    思考像是乾淨的水源、能源

  • you know, clean water, and clean energy,

    永續農業等問題

  • sustainable agriculture,

    並用它們無邊的想像提出解決之道

  • and come up with creative solutions with their imaginations.

    這將能改變世界的歷史

  • That would make it difference in the history of the world.

    事實上呢,我們人類製造問題的速度太快了

  • We're creating problems very quickly actually.

    有些我們想到的解決辦法