返回首页

NorikoArai_2017-_机器人可以通过大学入学考试吗？_
Today, I'm going to talk about AI and us. 今天我要跟大家聊聊 「人工智慧与你我」,
AI researchers have always said that we humans do not need to worry, because only menial jobs will be taken over by machines. 人工智慧研究专家常说, 人类不需要担心, 因为机器只会取代那些 乏味枯燥的粗活。
menial:adj.卑微的;仆人的;适合仆人做的;n.仆人;住家佣工;下贱的人;
Is that really true? 是真的吗？
They have also said that AI will create new jobs, so those who lose their jobs will find a new one. 他们也说过人工智慧 可以创造新的工作, 因此那些失業的人还是可以再就業。
Of course. 当然。
But the real question is: 只是问题的症结是:
How many of those who may lose their jobs to AI will be able to land a new one, especially when AI is smart enough to learn better than most of us? 多少人因为人工智慧失業后 能够真的找到新的工作, 尤其当人工智慧已经成熟到 比大多数人都能更有效率地学习？
especially:adv.尤其;特别;格外;十分;
Let me ask you a question: 让我问你们一个问题:
How many of you think that AI will pass the entrance examination of a top university by 2020? 人工智慧将可以通过 顶尖大学的入学测验的人,请举手。
Not many. OK. 哇,好多。好的。
So some of you may say, "Of course, yes!" 相当多人会认为:「那当然！」
Now singularity is the issue . 现在「人工智慧的奇点」 是个热门话题。
singularity:n.奇异;奇点;突出;稀有; issue:n.重要议题;争论的问题;v.宣布;公布;发出;发行;
And some others may say, "Maybe, because AI already won against a top Go player." 也有很多人会说:「有可能, 人工智慧都已经打败过 世界顶尖的围棋高手了。」
And others may say, "No, never. Uh-uh." 也有一些人持不同看法: 「绝对不可能。嗯,就是这样。」
That means we do not know the answer yet, right? 这说明了我们其实还没有 真正的答案,不是吗？
So that was the reason why I started Todai Robot Project, making an AI which passes the entrance examination of the University of Tokyo, the top university in Japan. 这就是我启动 「东大机器人计画」的始末, 打造一个人工智慧机器人, 通过东京大学的入学考, 东大是全日本最顶尖的学府。
This is our Todai Robot. 为您介绍东大机器人。
And, of course, the brain of the robot is working in the remote server. 当然,它正接受远端伺服器遥控着。
remote:adj.偏远的;偏僻的;遥远的;久远的;
It is now writing a 600-word essay on maritime trade in the 17th century. 正在写一篇 600 字的论文, 阐述 17 世纪的海上贸易。
maritime:adj.海的;海事的;沿海的;海员的;
How does that sound? 听起来如何？
Why did I take the entrance exam as its benchmark ? 为什么当初要将入学考试 作为一个标竿？
benchmark:n.基准;标准检查程序;vt.用基准问题测试(计算机系统等);
Because I thought we had to study the performance of AI in comparison to humans, especially on the scales and expertise which are believed to be acquired only by humans and only through education. 因为我想,我们有必要研究人工智慧 相较人类的表现, 尤其那些有规模 跟特殊性的专长领域, 向来我们都相信 这些技能和知识惟有人类 透过教育才能获得。
performance:n.性能;表现;业绩;表演; comparison:n.比较;对比;相比; scales:n.规模范围; v.攀登; (scale的第三人称单数和复数) expertise:n.专门知识;专门技术;专家的意见; acquired:adj.习得的; v.获得; (acquire的过去分词和过去式)
To enter Todai, the University of Tokyo, you have to pass two different types of exams. 要考上日本的 第一学府,东京大学, 必须通过两项测验。
The first one is a national standardized test in multiple-choice style. 第一个是日本高考, 全国标准化的测验, 全选择题的题型。
standardized:adj.标准的; v.使合乎标准; multiple-choice:adj.多项选择的;
You have to take seven subjects and achieve a high score -- 你必须在七大科目中 都获得高分——
I would say like an 84 percent or more accuracy rate -- to be allowed to take the second stage written test prepared by Todai. 大概要达到 85% 以上的正确率—— 才可以获准进入第二阶段笔试, 由东大命题。
accuracy:n.[数]精确度,准确性;
So let me first explain how modern AI works, taking the " Jeopardy !" challenge as an example. 先让我说明一下 当代的人工智慧运作的方式, 举一个 「Jeopardy!」 《危险边缘》 的例子。
Jeopardy:n.危险;(被告处于被判罪或受处罚的)危险境地;
Here is a typical "Jeopardy!" question: "Mozart's last symphony shares its name with this planet." 典型的一道题目如下: 「莫札特最后创作的交响曲 跟这个星球同名。」
typical:adj.典型的;特有的;象征性的; symphony:n.交响乐;谐声,和声;
Interestingly , a "Jeopardy!" question always asks, always ends with "this" something: "this" planet, "this" country, "this" rock musician, and so on. 有趣的是, 「这个」字眼在里头: 「这个」星球、「这个」国家, 「这个」摇滚乐手 或「这个」什么什么。
Interestingly:adv.有趣地;
In other words, "Jeopardy!" doesn't ask many different types of questions, but a single type, which we call " factoid questions." 也就是说,《危险边缘》 没有太多类型的题目, 几乎只有这一类, 我们称作「趣味小问题」。
factoid:n.仿真陈述(仅因出现在出版物上而被信以为真);adj.虚构的;似是而非的;
By the way , do you know the answer? 有没有人刚好知道答案呀？
By the way:顺便说一下;
If you do not know the answer and if you want to know the answer, what would you do? 假如我们不知道答案, 但是又想要知道, 怎么办？
You Google , right? Of course. 当然是上网搜寻呀——
Google:谷歌;谷歌搜索引擎;
Why not? 为甚么不呢?
But you have to pick appropriate keywords like "Mozart," "last" and "symphony" to search. 但要挑对关键字, 譬如输入像「莫札特」、 「最后」跟「交响曲」去搜寻。
appropriate:adj.适当的;恰当的;v.占用,拨出; keywords:n.关键词;关键字;
The machine basically does the same. 基本上机器人也是这么做。
basically:adv.主要地,基本上;
Then this Wikipedia page will be ranked top. 维基百科的页面就会出现在最上头。
Wikipedia:维基百科;
Then the machine reads the page. 所以机器人就开始 「读」这个页面吗？
No, uh-uh. 并不是喔──
Unfortunately , none of the modern AIs, including Watson, Siri and Todai Robot, is able to read. 不幸的是,所有当代的智慧机器人, 无论是 IBM 的华生、苹果的 Siri 或是东大机器人, 它们都没有「阅读」的能力。
Unfortunately:adv.不幸地; Siri:n.iPhone4S上的语音控制功能;
But they are very good at searching and optimizing . 但是它们在搜寻跟 得到最佳化结果上很在行。
optimizing:n.[数]优化,最佳化;adj.最佳的;v.最佳化(optimize的现在分词);
It will recognize that the keywords "Mozart," "last" and "symphony" 它会找到关键字像是 「莫札特」、「最后」 跟「交响曲」,
recognize:v.认识;认出;辨别出;承认;意识到;
are appearing heavily around here. 重复地出现在这一带。
So if it can find a word which is a planet and which is co-occurring with these keywords, that must be the answer. 接着继续寻找属于星球的词彙, 是跟前述这些关键字同时出现的, 那铁定就是答案了。
This is how Watson finds the answer " Jupiter ," in this case. 华生就是这样找到「木星」的。
Jupiter:n.[天]木星;朱庇特(罗马神话中的宙斯神);
Our Todai Robot works similarly , but a bit smarter in answering history yes-no questions, like, "'Charlemagne repelled the Magyars .' Is this sentence true or false?" 东大机器人的运作方式很接近, 但在回答历史科目的 判断题上表现稍好, 例如: 查理曼大帝击败马札尔人,
similarly:adv.同样地;类似于; repelled:v.击退;驱逐;赶走;使恐惧;使厌恶(repel的过去分词和过去式) Magyars:adj.匈牙利人的;马札尔人的;n.马札儿语;马札儿人;
Our robot starts producing a factoid question, like: "Charlemagne repelled [this person type]" by itself. 机器人自动转换为一道 趣味小问题, 变成:「 查理曼大帝 击败了这一种人」。
Then, "Avars" but not "Magyars" is ranked top. 结果最上头出现了「阿瓦尔人」 而非「马札尔人」。
This sentence is likely to be false. 所以这个陈述句很可能是错误的。
Our robot does not read, does not understand, but it is statistically correct in many cases. 机器人不会阅读,也不了解, 但从统计学角度评估, 却具有高准确度。
statistically:adv.统计地;统计学上;
For the second stage written test, it is required to write a 600-word essay like this one: 至于第二阶段的笔试, 受测者必须写一篇 600 字的论文,
[Discuss the rise and fall of the maritime trade in East and Southeast Asia in the 17th century ...] （阐述 17 世纪时东亚与东南亚 海上贸易的兴衰……） 如同我稍早展示过的,
Southeast:adj.东南的;来自东南的;n.东南;东南地区;adv.来自东南;
and as I have shown earlier, our robot took the sentences from the textbooks and Wikipedia, combined them together, and optimized it to produce an essay without understanding a thing. 我们的机器人将教科书 与维基百科的句子 併在一起, 优化后形成一篇文章, 完全不懂字里行间的意涵。
optimized:adj.最佳化的;尽量充分利用;
(Laughter) （笑声）
But surprisingly , it wrote a better essay than most of the students. 但是令人惊讶的是, 机器人这样写出来的文章, 居然比大多数的学生好。
surprisingly:adv.令人惊讶地;出乎意料地
(Laughter) （笑声）
How about mathematics ? 那么数学呢？
mathematics:n.数学;数学运算;
A fully automatic math-solving machine has been a dream since the birth of the word " artificial intelligence ," 能全自动处理数学问题的机器人, 是大家都梦寐以求的, 打从「人工智慧」的概念 问世以来就是如此。
automatic:adj.自动的;无意识的;必然的;n.自动步枪;自动换挡汽车; artificial intelligence:n.人工智能;
but it has stayed at the level of arithmetic for a long, long time. 但是,它曾经 长期停滞在算术的阶段。
arithmetic:n.算术,算法;
Last year, we finally succeeded in developing a system which solved pre-university-level problems from end to end, like this one. 去年我们总算成功发展一套系统, 可以从头到尾地 解决中等教育程度的数学题目, 像这一题。
finally:adv.终于;最终;(用于列举)最后;彻底地;
This is the original problem written in Japanese, and we had to teach it 2,000 mathematical axioms and 8,000 Japanese words to make it accept the problems written in natural language . 原文是日文。 我们必须先教会机器人 2,000 个数学公理, 与 8,000 个日文字, 才能让机器人 看懂原文的数学题目。
original:n.原件;原作;原物;原型;adj.原始的;最初的;独创的;新颖的; mathematical:adj.数学的,数学上的;精确的; axioms:n.[数]公理;公设;原理(axiom的复数); natural language:n.自然语言(自然发展而成,并非人造);
And it is now translating the original problems into machine-readable formulas . 它现在正在翻译原来的题目 成为机器人的语言。
machine-readable:adj.[计]机器可读的;可用计算机处理的; formulas:n.[数]公式;配方;规则(formula的复数);
Weird , but it is now ready to solve it, I think. 很怪,不过应该可以开始计算了。
Weird:adj.奇怪的;奇异的;离奇的;n.命运;宿命;命运女神;
Go and solve it. 开始解题。
Yes! It is now executing symbolic computation . 没错！它正在进行符号运算。
executing:v.(尤指依法)处决,处死;实行;执行;实施;(execute的现在分词) symbolic:adj.象征的;符号的;使用符号的; computation:n.估计,计算;
Even more weird, but probably this is the most fun part for the machine. 更怪了, 但或许机器人会觉得这个才好玩。
(Laughter) （笑声）
Now it outputs a perfect answer, though its proof is impossible to read, even for mathematicians . 好了,它产出了一个完美的解。 尽管连数学家都证实了, 完全没有人看得懂。
outputs:n.[计]输出; v.生产; proof:n.证据;证实;adj.能抵御;可防护; mathematicians:n.[数]数学家(mathematician的复数形式);
Anyway, last year our robot was among the top one percent in the second stage written exam in mathematics. 无论如何,去年我们的机器人 在第二阶段的数学表现中 被归类在排名前 1% 高分的群组中。
(Applause) （掌声）
Thank you. 谢谢。
So, did it enter Todai? 所以它最终有没有考上东大呢？
No, not as I expected. 它并没有如预期的金榜题名。
Why? 为什么？
Because it doesn't understand any meaning. 因为它根本什么也不懂。
Let me show you a typical error it made in the English test. 让我展示一个在英文科的典型错误。
[Nate: We're almost at the bookstore. Just a few more minutes. （奈特:我们快到书店了, 再过几分钟就到了。
Sunil: Wait. ______ . Nate: Thank you! That always happens ...] 两个人在对话。
Two people are talking. 我们都明白发生了什麽——
For us, who can understand the situation --
[1. "We walked for a long time." 2. "We're almost there." （选项:1. 我们走了很久的路 2. 我们几乎快到了
3. "Your shoes look expensive." 4. "Your shoelace is untied ."] 很明显地,标准答案 是选 4,同意吗？
shoelace:n.鞋带; untied:v.解开…的结;打开;(untie的过去分词和过去式)
it is obvious number four is the correct answer, right? 可是东大机器人选 2,
obvious:adj.明显的;显著的;平淡无奇的;
But Todai Robot chose number two, even after learning 15 billion English sentences using deep learning technologies . 就算已经学习了 150 亿个英文词句, 还透过深度学习技术。
technologies:n.技术;科技(technology的复数);
OK, so now you might understand what I said: modern AIs do not read, do not understand. 好吧,现在你可能明白我所说的: 当代人工智慧没有办法阅读, 不能理解。
They only disguise as if they do. 它们只是佯装成什麽都懂。
disguise:v.掩饰;假装;隐瞒;n.伪装;假装;用作伪装的东西;
This is the distribution graph of half a million students who took the same exam as Todai Robot. 这个分布图, 代表跟东大机器人一起接受入学 考试的其他 50 万名考生的成绩。
distribution:n.分布;分配;分发;分销;
Now our Todai Robot is among the top 20 percent, and it was capable to pass more than 60 percent of the universities in Japan -- but not Todai. 机器人排名其中的前 20%, 可以考进日本 超过六成的大学── 但就是考不上东大。
capable:adj.能干的,能胜任的;有才华的;
But see how it is beyond the volume zone of to-be white-collar workers. 可是看看被它超越的广大区块, 所谓的白领阶级。
volume:n.体积;容积;音量;响度;一册;合订本 to-be:adj.未来的;即将的; white-collar:adj.白领阶层的;脑力劳动者;
You might think I was delighted . 你可能会猜想我应该很开心。
delighted:adj.高兴的; v.使高兴; (delight的过去分词和过去式)
After all, my robot was surpassing students everywhere. 毕竟我的机器人正在 全面性地超越学生。
surpassing:adj.胜过的;卓越的;优秀的;adv.卓越地;非凡地;v.胜过(surpass的ing形式);
Instead, I was alarmed. 其实不然,我很惊恐。
How on earth could this unintelligent machine outperform students -- our children? 这个一点都不聪明的机器人居然 表现得比学生们── 也是我们的孩子们,更好？
unintelligent:adj.无知的;缺乏才智的; outperform:vt.胜过;做得比…好;
Right? 怎么可以？
I decided to investigate what was going on in the human world. 我决定深入调查 人类世界究竟发生了什么事。
investigate:v.调查;研究;审查;
I took hundreds of sentences from high school textbooks and made easy multiple-choice quizzes , and asked thousands of high school students to answer. 我收集了上百个 高中教科书里头的词句, 然后编成简单的选择题测验, 让上千位高中生接受测试。
quizzes:n.小测验(quiz复数形式);智力比赛;v.测验;盘问(quiz的第三人称单数形式);
Here is an example: 这是其中一个范例:
[Buddhism spread to ... , Christianity to ... and Oceania, and Islam to ...]
Christianity:n.基督教;基督教精神,基督教教义;
Of course, the original problems are written in Japanese, their mother tongue .
mother tongue:n.母语;本国语;
[ ______ has spread to Oceania.
Obviously, Christianity is the answer, isn't it?
It's written!
And Todai Robot chose the correct answer, too. 东大机器人也选出了正确的答案。
But one-third of junior high school students failed to answer this question. 但是有三分之一的国中生 无法回答这个问题。
junior high school:初中;
Do you think it is only the case in Japan? 你以为这个问题只存在于日本吗？
I do not think so, because Japan is always ranked among the top in OECD PISA tests, measuring 15-year-old students' performance in mathematics, science and reading every three years. 我不这么认为, 日本总是在国际学生能力评估计划 测验中名列前茅。 那是一套衡量 15 岁青少年在数学、 科学与阅读素质的测验, 每三年考一次。
PISA:n.比萨(意大利城市);
We have been believing that everybody can learn and learn well, as long as we provide good learning materials free on the web so that they can excel through the internet. 我们一直相信, 每个人都能学习, 并且学得出色, 只要我们提供高质量的学习资料, 这些免费资源, 让他们透过网路取得使用。
as long as:conj.只要;长达;如果;既然; excel:v.超过;擅长;
But such wonderful materials may benefit only those who can read well, and the percentage of those who can read well may be much less than we expected. 但这些优质的资料 只会让那些能够有效阅读的人受益, 而能够有效阅读的人所占的比例, 可能远低于我们的预期。
percentage:n.百分比;百分率;利润的分成;提成;
How we humans will coexist with AI is something we have to think about carefully, based on solid evidence . 人类要如何与人工智慧共存 是我们要谨慎思考的课题, 客观考量各项可靠证据。
coexist:vi.共存;和平共处; evidence:n.证据,证明;迹象;明显;v.证明;
At the same time , we have to think in a hurry because time is running out. 同时,我们也要加紧脚步思考, 因为所剩时间不多了。
At the same time:同时;另一方面;与此同时;
Thank you. 谢谢。
(Applause) （掌声）
Chris Anderson: Noriko, thank you. 克里斯 · 安德森:纪子,谢谢你。
Noriko Arai: Thank you. 新井纪子:谢谢。
CA: In your talk, you so beautifully give us a sense of how AIs think, what they can do amazingly and what they can't do. 克里斯:您方才向我们 说明了人工智慧的运作方式。 它们可以完美胜任的, 以及无法胜任的工作。
But -- do I read you right, that you think we really need quite an urgent revolution in education to help kids do the things that humans can do better than AIs? 但是,我的解读是否是正确的, 你认为我们急需教育改革, 以协助学子在特定的领域中 让人工智慧难以望其项背？
urgent:adj.紧急的;急迫的; revolution:n.革命;旋转;运行;循环;
NA: Yes, yes, yes. 纪子:是的。
Because we humans can understand the meaning. 纪子:是的。
That is something which is very, very lacking in AI. 而这在人工智慧中是相当缺乏的。
But most of the students just pack the knowledge without understanding the meaning of the knowledge, so that is not knowledge, that is just memorizing , and AI can do the same thing. 但是大多数的学生都只会囫囵吞枣, 而非深入理解知识, 这就只是单纯记忆的动作, 人工智慧也办得到。
memorizing:n.记忆;熟记;v.熟记;纪念(memorize的现在分词);
So we have to think about a new type of education. 所以我们应该要思考新型态的教育。
CA: A shift from knowledge, rote knowledge, to meaning. 克里斯:从「死记硬背」 到「深入理解」的转变。
shift:n.移动;变化;手段;轮班;v.移动;转变;转换; rote:n.死记硬背;生搬硬套;
NA: Mm-hmm. 纪子:嗯嗯嗯。
CA: Well, there's a challenge for the educators. Thank you so much. 克里斯:我想这是给 教育家的一大挑战,
NA: Thank you very much. Thank you. 纪子:谢谢,非常谢谢你们。
(Applause) （掌声）