Mark Twain summed up what I take to be one of the fundamental problems of cognitive science with a single witticism. He said, "There's something fascinating about science. One gets such wholesale returns of conjecture out of such a trifling investment in fact." (Laughter)
馬克.吐溫用一句妙語概括了, 我認為是認知科學的一個最根本問題。 他說:「科學的有趣之處在於, 一個人可從微不足道的事得出了偉大的猜想。」 (笑聲)
Twain meant it as a joke, of course, but he's right: There's something fascinating about science. From a few bones, we infer the existence of dinosuars. From spectral lines, the composition of nebulae. From fruit flies, the mechanisms of heredity, and from reconstructed images of blood flowing through the brain, or in my case, from the behavior of very young children, we try to say something about the fundamental mechanisms of human cognition. In particular, in my lab in the Department of Brain and Cognitive Sciences at MIT, I have spent the past decade trying to understand the mystery of how children learn so much from so little so quickly. Because, it turns out that the fascinating thing about science is also a fascinating thing about children, which, to put a gentler spin on Mark Twain, is precisely their ability to draw rich, abstract inferences rapidly and accurately from sparse, noisy data. I'm going to give you just two examples today. One is about a problem of generalization, and the other is about a problem of causal reasoning. And although I'm going to talk about work in my lab, this work is inspired by and indebted to a field. I'm grateful to mentors, colleagues, and collaborators around the world.
馬克.吐溫當然只是開玩笑,但他是對的。 科學有其有趣之處。 從幾塊骨頭,我們推測了恐龍的存在; 從譜線得出了星雲的成份; 從果蠅得出了遺傳的機制; 以及從血液流入大腦的重建影像, 在我的研究則是從幼兒的行為中, 我們嘗試解釋人類認知的基本機制。 我在麻省理工大腦及認知科學系實驗室中, 花了過去十年研究一個謎團, 就是兒童如何從零開始, 快速地學到那麼多的東西。 科學令人著迷之處, 亦正是孩子令人著迷的地方。 回應馬克.吐溫的話, 那就是孩子從零碎和離亂的訊息中, 能夠得出豐富而抽象的推論的能力。 我將舉出兩個例子: 一個是關於廣義化的問題, 另一個則是關於因果推理的。 雖然我將會談及我實驗室的研究, 但這個研究的靈感是來自一個領域, 一個我要感謝世界各地的導師、 同事和工作夥伴付出的領域。
Let me start with the problem of generalization. Generalizing from small samples of data is the bread and butter of science. We poll a tiny fraction of the electorate and we predict the outcome of national elections. We see how a handful of patients responds to treatment in a clinical trial, and we bring drugs to a national market. But this only works if our sample is randomly drawn from the population. If our sample is cherry-picked in some way -- say, we poll only urban voters, or say, in our clinical trials for treatments for heart disease, we include only men -- the results may not generalize to the broader population.
讓我先談談廣義化的問題。 歸納數據樣本在科學上是不可或缺的, 如我們調查一部分的選民, 然後預測國家大選的結果。 我們觀察一小撮病人在臨床試驗中的反應, 然後把藥物帶入市場, 但只有在整個人口中隨機抽樣才可行。 當我們刻意挑選樣本, 如我們只調查城市中的選民, 又或在治療心臟病的臨床試驗中, 我們只研究男性, 這樣的結果便不能代表整個人口。
So scientists care whether evidence is randomly sampled or not, but what does that have to do with babies? Well, babies have to generalize from small samples of data all the time. They see a few rubber ducks and learn that they float, or a few balls and learn that they bounce. And they develop expectations about ducks and balls that they're going to extend to rubber ducks and balls for the rest of their lives. And the kinds of generalizations babies have to make about ducks and balls they have to make about almost everything: shoes and ships and sealing wax and cabbages and kings.
因此科學家著緊抽樣的方法是否隨機。 但這又跟嬰兒有甚麼關係? 嬰兒在任何時候都要歸納數據樣本, 當他們看到幾隻橡皮鴨, 並知道它們浮在水面。 又或見到幾個皮球, 並知道它們能彈跳。 從中他們建立對橡膠鴨和皮球的概念, 並將這概念延伸至日後會見到的 所有橡膠鴨和皮球。 嬰兒對橡膠鴨和皮球的這種概括, 他們會運用在每一件事上: 鞋子、船、封蠟、捲心菜和皇帝。
So do babies care whether the tiny bit of evidence they see is plausibly representative of a larger population? Let's find out. I'm going to show you two movies, one from each of two conditions of an experiment, and because you're going to see just two movies, you're going to see just two babies, and any two babies differ from each other in innumerable ways. But these babies, of course, here stand in for groups of babies, and the differences you're going to see represent average group differences in babies' behavior across conditions. In each movie, you're going to see a baby doing maybe just exactly what you might expect a baby to do, and we can hardly make babies more magical than they already are. But to my mind the magical thing, and what I want you to pay attention to, is the contrast between these two conditions, because the only thing that differs between these two movies is the statistical evidence the babies are going to observe. We're going to show babies a box of blue and yellow balls, and my then-graduate student, now colleague at Stanford, Hyowon Gweon, is going to pull three blue balls in a row out of this box, and when she pulls those balls out, she's going to squeeze them, and the balls are going to squeak. And if you're a baby, that's like a TED Talk. It doesn't get better than that. (Laughter) But the important point is it's really easy to pull three blue balls in a row out of a box of mostly blue balls. You could do that with your eyes closed. It's plausibly a random sample from this population. And if you can reach into a box at random and pull out things that squeak, then maybe everything in the box squeaks. So maybe babies should expect those yellow balls to squeak as well. Now, those yellow balls have funny sticks on the end, so babies could do other things with them if they wanted to. They could pound them or whack them. But let's see what the baby does.
因此嬰兒留意這些細節能否代表整體。 我們一起看看吧。 我將讓你看兩段短片, 這兩段短片分別代表實驗的兩個情況。 由於你將看到兩段短片, 你只會看到兩個嬰兒, 而這兩個嬰兒在很多地方都是不同的。 但這兩個嬰兒將代表更大的群組, 你將看到的不同之處則代表 嬰兒行為中的平均差異。 在每一段短片中你會見到嬰兒 在做些他們正常會做的事。 嬰兒本身已是十分神奇的, 但對我來說他們的神奇之處, 也是我想你們留意的地方, 就是這兩種情況之間的分別。 因為這兩段短片唯一不同的地方, 正是嬰兒將要觀察的資料。 我們把一些藍色和黃色的球給嬰兒看。 權孝媛當時是我的學生, 現在則是史丹佛大學的同事。 她將拿出三個藍色的球, 而當她拿出這些球時, 她會把球擠一下, 讓這些球發出吱吱聲。 這對於嬰兒來說就像TED一樣, 是件很美好的事。 (笑聲) 從一個裝滿藍色球的箱中, 抽出三個藍色球是件很容易的事。 你閉上眼睛也能做到, 這就像隨機抽樣。 因此當你可以在箱中隨機地抽出 能吱吱叫的物件時, 也許箱中所有物件都能吱吱叫, 所以嬰兒可能會假設黃色球也能吱吱叫。 但這些黃色球都有一根棒, 所以嬰兒可用它們做些不同的事, 他們可以拍打或搖動這些球。 就讓我們看看這嬰兒會做甚麼。
(Video) Hyowon Gweon: See this? (Ball squeaks) Did you see that? (Ball squeaks) Cool. See this one? (Ball squeaks) Wow.
(影片) 權孝媛: 看看這個。 (球發出吱吱聲) 看到這個嗎? (球發出吱吱聲) 很酷吧! 看看這個。 (球發出吱吱聲) 哇!
Laura Schulz: Told you. (Laughs)
羅拉·舒爾茨: 早就說了。 (笑聲)
(Video) HG: See this one? (Ball squeaks) Hey Clara, this one's for you. You can go ahead and play. (Laughter)
(影片) 孝媛: 看到這個嗎? (球發出吱吱聲) 克拉拉, 這個是給你的, 你拿去玩吧。 (笑聲)
LS: I don't even have to talk, right? All right, it's nice that babies will generalize properties of blue balls to yellow balls, and it's impressive that babies can learn from imitating us, but we've known those things about babies for a very long time. The really interesting question is what happens when we show babies exactly the same thing, and we can ensure it's exactly the same because we have a secret compartment and we actually pull the balls from there, but this time, all we change is the apparent population from which that evidence was drawn. This time, we're going to show babies three blue balls pulled out of a box of mostly yellow balls, and guess what? You [probably won't] randomly draw three blue balls in a row out of a box of mostly yellow balls. That is not plausibly randomly sampled evidence. That evidence suggests that maybe Hyowon was deliberately sampling the blue balls. Maybe there's something special about the blue balls. Maybe only the blue balls squeak. Let's see what the baby does.
羅拉: 我不用解釋, 對吧? 嬰兒把藍色球的特性套用到黃色球上。 嬰兒從模仿我們中學習,這是很神奇的, 但我們早就知道嬰兒能這樣做。 有趣的地方是當把一樣的東西給嬰兒看時, 甚麼事會發生。 我們能肯定這是完全一樣的, 因我們有個秘密的空間, 從中我們抽出這些球。 但這次我們改變了抽樣的母體。 這次我們在一個裝滿黃色球的箱中, 抽出三個藍色球給嬰兒看。 想想甚麼事會發生? 你大概不能隨機地在裝滿黃色球的箱中, 連續抽出三個藍色球, 因此這很可能不是隨機抽樣。 這反映了孝媛可能是刻意抽出藍色球, 可能這些藍色球是特別的, 可能只有藍色球能吱吱叫。 一起看看這嬰兒會做甚麼。
(Video) HG: See this? (Ball squeaks) See this toy? (Ball squeaks) Oh, that was cool. See? (Ball squeaks) Now this one's for you to play. You can go ahead and play.
(影片) 孝媛: 看看這個。 (球發出吱吱聲) 看到這個玩具嗎? (球發出吱吱聲) 哇, 這很酷, 看到嗎? (球發出吱吱聲) 這個是給你的, 你拿去玩吧。
(Fussing) (Laughter)
(不耐煩的) (笑聲)
LS: So you just saw two 15-month-old babies do entirely different things based only on the probability of the sample they observed. Let me show you the experimental results. On the vertical axis, you'll see the percentage of babies who squeezed the ball in each condition, and as you'll see, babies are much more likely to generalize the evidence when it's plausibly representative of the population than when the evidence is clearly cherry-picked. And this leads to a fun prediction: Suppose you pulled just one blue ball out of the mostly yellow box. You [probably won't] pull three blue balls in a row at random out of a yellow box, but you could randomly sample just one blue ball. That's not an improbable sample. And if you could reach into a box at random and pull out something that squeaks, maybe everything in the box squeaks. So even though babies are going to see much less evidence for squeaking, and have many fewer actions to imitate in this one ball condition than in the condition you just saw, we predicted that babies themselves would squeeze more, and that's exactly what we found. So 15-month-old babies, in this respect, like scientists, care whether evidence is randomly sampled or not, and they use this to develop expectations about the world: what squeaks and what doesn't, what to explore and what to ignore.
羅拉: 你剛剛看到兩個15月大的嬰兒, 按他們觀察到樣本出現的機率, 而做出完全不同的事。 一起看看實驗的結果, 垂直軸代表在每一個情況中, 有多少百分比的嬰兒擠壓球。 你可看見嬰兒在樣本和整體一致時, 比刻意挑選的樣本, 較會歸納他們看到的特徵。 因此這帶出一個有趣的預測。 假設你在一個裝滿黃色球的箱中 只拿出一個藍色球, 當然你很難隨機地連續抽出三個藍色球, 但你可以只用一個藍色球作樣本, 這不一定是個不可行的樣本。 當你隨機抽出一個會吱吱叫的東西時, 可能箱中所有的東西都會吱吱叫, 因此雖然嬰兒會看到較少吱吱叫的例子, 而且在只抽出一個球的情況下, 他們會有較少的動作去模仿, 但我們預計會有更多嬰兒擠壓球。 這正是我們發現的結果。 因此15月大的嬰兒在這方面就像科學家, 他們留意抽樣的方法是否隨機, 並以此建立對事物的概念: 甚麼會吱吱叫而甚麼不會, 甚麼需要探索而甚麼可忽略。
Let me show you another example now, this time about a problem of causal reasoning. And it starts with a problem of confounded evidence that all of us have, which is that we are part of the world. And this might not seem like a problem to you, but like most problems, it's only a problem when things go wrong. Take this baby, for instance. Things are going wrong for him. He would like to make this toy go, and he can't. I'll show you a few-second clip. And there's two possibilities, broadly: Maybe he's doing something wrong, or maybe there's something wrong with the toy. So in this next experiment, we're going to give babies just a tiny bit of statistical data supporting one hypothesis over the other, and we're going to see if babies can use that to make different decisions about what to do.
現在讓我給你們看看另一個例子, 這次是關於因果推理的。 每人都要面對這個問題, 因為我們都是這世界的一部份。 這看似不是一個問題, 但和其他問題一樣, 事情會出狀況。 以這個嬰兒為例, 所有事都出了問題, 他想開動這個玩具,但他做不到。 我會讓你看一段幾秒的影片。 這有兩個可能的原因, 可能是他做錯了一些事, 又或是那個玩具有些問題。 因此在這個實驗中, 我們會給嬰兒們少許資料。 這些資料會傾向支持其中一個可能性, 我們將研究這些嬰兒能否運用這些資料, 而作出不同的決定。
Here's the setup. Hyowon is going to try to make the toy go and succeed. I am then going to try twice and fail both times, and then Hyowon is going to try again and succeed, and this roughly sums up my relationship to my graduate students in technology across the board. But the important point here is it provides a little bit of evidence that the problem isn't with the toy, it's with the person. Some people can make this toy go, and some can't. Now, when the baby gets the toy, he's going to have a choice. His mom is right there, so he can go ahead and hand off the toy and change the person, but there's also going to be another toy at the end of that cloth, and he can pull the cloth towards him and change the toy. So let's see what the baby does.
這個實驗是這樣的: 孝媛嘗試開動那個玩具並成功了, 而我的兩次嘗試都失敗了, 之後孝媛再嘗試,並再次成功了。 這就像我和我的學生在使用新科技的情況。 重要的是這提供了少許的資料, 這反映玩具並沒有問題,而是人的問題。 有些人可以開動這玩具, 有些人則不能。 當這嬰兒拿到玩具時,他要作一個選擇。 他的母親在旁, 所以他可以把玩具交給母親, 換另一人嘗試。 同時在毛巾上有另一個玩具, 所以他也可以把玩具拉向自己, 換另一個玩具。 一起看看嬰兒會怎樣做。
(Video) HG: Two, three. Go! (Music) LS: One, two, three, go! Arthur, I'm going to try again. One, two, three, go! YG: Arthur, let me try again, okay? One, two, three, go! (Music) Look at that. Remember these toys? See these toys? Yeah, I'm going to put this one over here, and I'm going to give this one to you. You can go ahead and play. LS: Okay, Laura, but of course, babies love their mommies. Of course babies give toys to their mommies when they can't make them work. So again, the really important question is what happens when we change the statistical data ever so slightly. This time, babies are going to see the toy work and fail in exactly the same order, but we're changing the distribution of evidence. This time, Hyowon is going to succeed once and fail once, and so am I. And this suggests it doesn't matter who tries this toy, the toy is broken. It doesn't work all the time. Again, the baby's going to have a choice. Her mom is right next to her, so she can change the person, and there's going to be another toy at the end of the cloth. Let's watch what she does.
(影片) 孝媛: 二、三、開始! (音樂) 羅拉: 一、二、三、開始! 亞瑟,讓我再試一次, 一、二、三、開始! 孝媛: 亞瑟,讓我再試吧,好嗎? 一、二、三、開始! (音樂) 看看這裡,記得這些玩具嗎? 看到嗎? 對,我會把這個放在這裡, 把另一個給你。 你拿去玩吧。 羅拉: 你或許會說嬰兒都愛他們的母親, 因此當玩具出現問題時, 嬰兒自然會把它交給母親。 因此,問題在於當我們稍微改變資料時, 甚麼事會發生。 這次,嬰兒將看到這玩具 按同一次序成功運作和失敗, 但我們改變了資料的分佈。 這次孝媛和我各有一次成功和一次失敗, 這代表誰人嘗試都沒有分別, 那件玩具是壞的, 它不是每次都能運作的。 同樣地,嬰兒要作出一個選擇, 她的母親在旁,所以她可換另一人嘗試, 同時另一個玩具就在毛巾上。 看看她會怎樣做。 (影片) 孝媛: 二、三、開始! (音樂)
(Video) HG: Two, three, go! (Music) Let me try one more time. One, two, three, go! Hmm.
讓我再試一次, 一、二、三、開始! 嗯...
LS: Let me try, Clara. One, two, three, go! Hmm, let me try again. One, two, three, go! (Music) HG: I'm going to put this one over here, and I'm going to give this one to you. You can go ahead and play. (Applause)
羅拉: 讓我試試吧,克拉拉。 一、二、三、開始! 嗯...讓我再試試。 一、二、三、開始! (音樂) 孝媛: 我把這個放在這裡, 這個則交給你, 你拿去玩吧。 (掌聲)
LS: Let me show you the experimental results. On the vertical axis, you'll see the distribution of children's choices in each condition, and you'll see that the distribution of the choices children make depends on the evidence they observe. So in the second year of life, babies can use a tiny bit of statistical data to decide between two fundamentally different strategies for acting in the world: asking for help and exploring. I've just shown you two laboratory experiments out of literally hundreds in the field that make similar points, because the really critical point is that children's ability to make rich inferences from sparse data underlies all the species-specific cultural learning that we do. Children learn about new tools from just a few examples. They learn new causal relationships from just a few examples. They even learn new words, in this case in American Sign Language.
羅拉: 看看這個實驗的結果, 在垂直軸上,你會看到在每種情況下, 嬰兒作出不同選擇的分佈。 你會發現他們作的選擇是 基於他們觀察到的資料。 因此當他們兩歲時, 嬰兒已經可以運用細微的資料, 在兩個完全不同的選項中作出決定: 尋求幫忙或自行探索。 我剛才讓你們看了兩個實驗, 在這領域中有數千個得出相同結果的實驗。 當中反映的重點是, 兒童擁有充分解讀零碎資訊的能力, 這超出了所有文化的學習方式。 孩子從少數的例子便能學到新技能, 他們從少數的例子便能領略到新的因果關係, 他們甚至能學到新的生字,如美國手語。
I want to close with just two points. If you've been following my world, the field of brain and cognitive sciences, for the past few years, three big ideas will have come to your attention. The first is that this is the era of the brain. And indeed, there have been staggering discoveries in neuroscience: localizing functionally specialized regions of cortex, turning mouse brains transparent, activating neurons with light. A second big idea is that this is the era of big data and machine learning, and machine learning promises to revolutionize our understanding of everything from social networks to epidemiology. And maybe, as it tackles problems of scene understanding and natural language processing, to tell us something about human cognition. And the final big idea you'll have heard is that maybe it's a good idea we're going to know so much about brains and have so much access to big data, because left to our own devices, humans are fallible, we take shortcuts, we err, we make mistakes, we're biased, and in innumerable ways, we get the world wrong. I think these are all important stories, and they have a lot to tell us about what it means to be human, but I want you to note that today I told you a very different story. It's a story about minds and not brains, and in particular, it's a story about the kinds of computations that uniquely human minds can perform, which involve rich, structured knowledge and the ability to learn from small amounts of data, the evidence of just a few examples. And fundamentally, it's a story about how starting as very small children and continuing out all the way to the greatest accomplishments of our culture, we get the world right.
我會提出兩個重點作總結。 如果你近年有留意大腦和認知科學領域, 你會聽到三個重要的概念。 第一,現在是大腦的時代。 的確,神經科學近來有不少驚人的發現, 例如標記了大腦皮層負責不同功能的位置、 製造出透明的老鼠大腦、 以及利用光線啟動神經元。 第二個重要的概念是, 現在是大數據和機器學習的時代, 而機器學習能徹底改變我們對任何事的理解, 從社交網站到流行病學。 當機器學習能理解埸合和處理自然語言時, 也許我們能藉此了解人類的認知。 最後一個你會聽過的重要概念是, 我們將對大腦有很深入的認識, 並能掌握大數據,而這很可能是件好事。 因為相比機器而言, 人類易犯錯誤,我們會走捷徑, 我們會做錯, 我們在很多方面都有偏見, 我們會有錯誤的理解。 我認為這都是重要的, 因為這反映了人類的特質, 但我今天想帶出事情的另一面。 這是關於思維而非大腦的, 尤其是人類獨有的運算能力, 這牽涉了豐富、有條理的知識, 以及從少量的數據和例子中學習的能力。 再者,這是關於我們如何從幼童, 一路發展至成為文化中偉大的成就, 我們能正確地理解這個世界。
Folks, human minds do not only learn from small amounts of data. Human minds think of altogether new ideas. Human minds generate research and discovery, and human minds generate art and literature and poetry and theater, and human minds take care of other humans: our old, our young, our sick. We even heal them. In the years to come, we're going to see technological innovations beyond anything I can even envision, but we are very unlikely to see anything even approximating the computational power of a human child in my lifetime or in yours. If we invest in these most powerful learners and their development, in babies and children and mothers and fathers and caregivers and teachers the ways we invest in our other most powerful and elegant forms of technology, engineering and design, we will not just be dreaming of a better future, we will be planning for one.
大家, 人腦不只是懂得從少量的數據中學習。 人腦能想到新的主意。 人腦能創造出研究和發明。 人腦能創作藝術、文學、寫詩和戲劇。 人腦可照顧其他人, 包括年老的、年輕的、患病的, 我們甚至能治癒他們。 在未來,我們將會看到 超乎現在能想像的科技發展, 但在我或你們的一生中, 我們不太可能目睹比得上嬰兒運算能力的機器。 假如我們投資在最厲害的學習者和其發展身上, 在嬰兒和兒童身上、 在母親和父親身上、 在照顧者和老師身上, 如同我們投資在最厲害的科技、工程和設計上時, 我們不只是夢想有個更好的將來, 而是在計劃一個更好的將來。 謝謝。
Thank you very much.
(掌聲)
(Applause)
Chris Anderson: Laura, thank you. I do actually have a question for you. First of all, the research is insane. I mean, who would design an experiment like that? (Laughter) I've seen that a couple of times, and I still don't honestly believe that that can truly be happening, but other people have done similar experiments; it checks out. The babies really are that genius.
克里斯·安德森: 羅拉, 謝謝你, 我其實想問你一個問題。 首先,這項研究真是太瘋狂了。 我的意思是,有誰會想到這些實驗? (笑聲) 我見過很多類似的實驗, 但我仍然覺得難以置信, 儘管很多人做了類似的實驗,而事實的確如此。 這些嬰兒根本是天才。
LS: You know, they look really impressive in our experiments, but think about what they look like in real life, right? It starts out as a baby. Eighteen months later, it's talking to you, and babies' first words aren't just things like balls and ducks, they're things like "all gone," which refer to disappearance, or "uh-oh," which refer to unintentional actions. It has to be that powerful. It has to be much more powerful than anything I showed you. They're figuring out the entire world. A four-year-old can talk to you about almost anything. (Applause)
羅拉: 在實驗中這看似很神奇, 但想想在現實生活中是怎樣的,對嗎? 一出世時,他只是個嬰兒, 但18個月後他開始說話, 而嬰兒最初說的話不只是物件, 如皮球和鴨子, 他們更能表達「不見了」的概念, 又或是以「哎喲」表達無心之失。 這必須是那麼厲害的, 這必須比我剛才展示的還要厲害。 嬰兒在弄清楚整個世界, 一個四歲的小孩幾乎懂得說所有東西。 (掌聲)
CA: And if I understand you right, the other key point you're making is, we've been through these years where there's all this talk of how quirky and buggy our minds are, that behavioral economics and the whole theories behind that that we're not rational agents. You're really saying that the bigger story is how extraordinary, and there really is genius there that is underappreciated.
克里斯: 如果我沒錯的話, 你想指出的另一個重點是, 這些年來,我們都聽說 我們的腦袋是不可信和會出錯的, 行為經濟學和其他新理論都指出我們不是理性的。 但你指出了我們的腦袋是非凡的, 我們一直忽略了我們的腦袋是多麼神奇。 羅拉: 我最喜愛的心理學名言之一,
LS: One of my favorite quotes in psychology comes from the social psychologist Solomon Asch, and he said the fundamental task of psychology is to remove the veil of self-evidence from things. There are orders of magnitude more decisions you make every day that get the world right. You know about objects and their properties. You know them when they're occluded. You know them in the dark. You can walk through rooms. You can figure out what other people are thinking. You can talk to them. You can navigate space. You know about numbers. You know causal relationships. You know about moral reasoning. You do this effortlessly, so we don't see it, but that is how we get the world right, and it's a remarkable and very difficult-to-understand accomplishment.
來自社會心理學家所羅門·阿希, 他說心理學首要的任務是 去除那些毋需証明的事物的面紗。 每天你作出大大小小的決定去理解這個世界。 你知道不同物件及其特性, 即使被覆蓋和在黑暗中你也知道。 你能在空間中行走。 你能猜到別人在想甚麼,你能和別人交談。 你能探索空間,你明白數字。 你明白因果關係, 你懂得分辨是非。 你毫不費力便能做到, 所以我們不會察覺, 但這就是我們理解這個世界的方法, 這是個神奇而又難以理解的成就。
CA: I suspect there are people in the audience who have this view of accelerating technological power who might dispute your statement that never in our lifetimes will a computer do what a three-year-old child can do, but what's clear is that in any scenario, our machines have so much to learn from our toddlers. LS: I think so. You'll have some machine learning folks up here. I mean, you should never bet against babies or chimpanzees or technology as a matter of practice, but it's not just a difference in quantity, it's a difference in kind. We have incredibly powerful computers, and they do do amazingly sophisticated things, often with very big amounts of data. Human minds do, I think, something quite different, and I think it's the structured, hierarchical nature of human knowledge that remains a real challenge.
克里斯: 我相信在坐有人認為科技正急速發展, 他們可能不認同你說電腦 不能做到三歲小孩能做到的事。 但可以肯定的是,無論在甚麼場合, 嬰兒有很多地方值得我們的機器學習。 羅拉: 我同意。有些人認同機器學習。 我的意思是,你不應將嬰兒和黑猩猩跟科技比較, 因為這不是數量上的不同, 而是性質上的不同。 我們有十分厲害的電腦, 它們能做到複雜的事情, 和處理大量的資料。 我認為人類的腦袋做的事是不同的, 人類的知識是有系統和條理分明的, 這對機器仍然是一個挑戰。 克里斯: 勞拉·舒爾茨,十分精彩。謝謝。
CA: Laura Schulz, wonderful food for thought. Thank you so much.
羅拉: 謝謝。 (掌聲)
LS: Thank you. (Applause)