视频库 / REC_016
--:--
点击播放，这里会跟随视频显示当前句的中英字幕。
第 16 期 · 回应 Ⅰ·11「人类还是特殊的吗？」
Why This Is the Most Exciting Time to Be Human | Ken Ono, Axiom Math
EO章节 · 点击跳转视频
0:00
开场：出不倒的 ChatGPT 难题
1:07
FrontierMath 与崩溃时刻
2:56
史上最强图书馆员的比喻
4:08
重新定义智力：创造与迁移
5:57
数学家之子的叛逆少年
7:02
拉马努金遗孀的一封信
8:40
父亲的战后人生与希望
10:23
追随拉马努金的学术转折
11:24
寻找散落人间的拉马努金
13:13
打勾式教育的批判
15:31
知识廉价后大学剩下什么
16:50
守护惊奇感与身份所有权
本期讲者
小野健美国数论学家，弗吉尼亚大学教授（现休假），拉马努金研究与整数分拆领域的代表人物，曾任电影《知无涯者》数学顾问。2025 年起担任 AI 初创公司 Axiom Math 的创始数学家。
主持人韩国创业媒体 EO 的访谈者，片中仅以简短回应出现，全片以 Ono 的第一人称叙述为主。
01开场：出不倒的 ChatGPT 难题
0:00
Exactly 1 year ago, I was happy-go-lucky university professor writing my papers. I would have described myself as someone who had enjoyed the privilege of being good at mathematics. And then there was a dramatic change. Now, I'm deeply troubled. For the first time, I struggled to assemble questions that ChatGPT would get wrong. These models know more facts than any human you would ever find. I was devastated. I was thinking, "How am I going to stay ahead of AI?" Is that I actually think that's the wrong question. Knowledge quickly became cheap. If our goal is to always stay ahead of AI, then I think we're going to lose.
就在一年前，我还是个无忧无虑的大学教授，写着我的论文。我会形容自己是那种有幸擅长数学的人。然后，发生了戏剧性的变化。现在，我深感不安。第一次，我很难出得出让 ChatGPT 答错的题目。这些模型知道的事实，比你能找到的任何人都多。我崩溃了。我当时在想："我要怎么才能领先于 AI？"但我其实觉得，这个问题问错了。知识很快就变得廉价了。如果我们的目标是永远领先于 AI，那我觉得我们注定要输。
0:47
My name is Ken Ono. I'm a mathematician, and I also work in the space called AI for math. I'm a professor at the University of Virginia on leave, and I'm the founding mathematician at Axiom Math.
我叫 Ken Ono。我是一名数学家，同时也在所谓的"AI for math"（人工智能助力数学）这个领域工作。我目前在弗吉尼亚大学任教授，不过在休假中，同时我是 Axiom Math 的创始数学家。
02FrontierMath 与崩溃时刻
1:07
The very first time I heard the term artificial intelligence was in 1993, and I met a faculty member who said, "I work in artificial intelligence." My very first words [music] were, "Oh, that's interesting. Um I specialize in natural intelligence." And I was cocky. I would have my butt handed to me many times for using words like that, but that's where I came from. Exactly 1 year ago, I was happy-go-lucky university professor writing my papers, enjoying life at the University of Virginia. And then there was a dramatic change. I came face-to-face with large language models at work, the Frontier Math program, where this company based in Berkeley, called Epoch AI, hired professional mathematicians from around the world to assemble very difficult math problems. And their goal was to assess the capabilities of large language models as as they improve. For the first time, I struggled to assemble questions that ChatGPT would get wrong.
我第一次听到"人工智能"这个词是在 1993 年，当时我遇到一位教员，他说："我是做人工智能的。"我脱口而出的第一句话【音乐】是："哦，那挺有意思的。呃，我专攻的是自然智能。"我当时挺自大的。后来我因为说这种话被狠狠打过很多次脸，但我就是那么一路走过来的。就在一年前，我还是个无忧无虑的大学教授，写着我的论文，在弗吉尼亚大学享受生活。然后就发生了戏剧性的变化。我在工作中正面接触到了大语言模型，那是 Frontier Math项目——伯克利的一家公司 Epoch AI，雇了全世界的职业数学家来出非常难的数学题。他们的目标是评估大语言模型随着迭代不断提升的能力。第一次，我很难出得出让 ChatGPT 答错的题目。
2:17
I was devastated. At the time, I was one of the few scientists who had been given access [music] to these state-of-the-art models, and I had to be patient for a few months to go by before the world did recognize that [music] yeah, these models know so much. These models know more facts than any human you would ever find. So, for a few months, I was devastated. I was thinking, "How am I going to stay [music] ahead of AI?" But I actually think that's the wrong question. If our goal is to always stay ahead of AI, then uh I think we're going to lose.
我崩溃了。当时，我是少数几个被授予【音乐】这些最先进模型使用权限的科学家之一，我不得不耐心等了好几个月，世界才终于意识到【音乐】——是的，这些模型懂得太多了。这些模型知道的事实，比你能找到的任何人都多。所以，有那么几个月，我整个人是崩溃的。我当时在想："我要怎么才能【音乐】领先于 AI？"但我其实觉得，这个问题问错了。如果我们的目标是永远领先于 AI，那呃我觉得我们注定要输。
03史上最强图书馆员的比喻
2:56
Nobody would be interested in watching Usain Bolt race against a motorcycle in the 1-mile run. It's not a fair race. But we still watch the Olympics. We, as a society, now know how to accept that machines can outperform humans in every physical way, but we're still now coming to grips with the fact that the brain, deep inquiry, computers have caught up. The large language models should be thought of as the most extraordinary librarian the world has ever seen. If [music] it has been written down, the large language model has probably seen it. If it's on YouTube, large language model has probably been trained on it.
没人会有兴趣看博尔特跟一辆摩托车比一英里赛跑。那不是一场公平的比赛。但我们还是会看奥运会。作为一个社会，我们现在已经懂得接受机器在各种体能方面都能超越人类，但我们还在慢慢消化这样一个事实：在大脑、在深度思考这件事上，计算机已经追上来了。大语言模型应该被看作是这世界上有史以来最了不起的图书馆员。如果【音乐】某样东西被写下来过，大语言模型很可能已经看过了。如果它在 YouTube 上，大语言模型很可能已经拿它训练过了。
3:37
If you read a newspaper article, by the afternoon, the large language model has probably seen it. Good luck with competing [music] with that ability to collect information. It Information, knowledge is now cheap. But how you use it and how you verify it has become more expensive. Do you want your librarian to be your neurosurgeon? >> [music] >> Do you want your librarian to be your air traffic controller somehow keeping an eye on the hundreds of planes that are flying over North America or Korea?
如果你读了一篇报纸文章，到了下午，大语言模型很可能已经看过它了。祝你好运，去跟那种收集信息的能力[音乐]竞争。信息、知识现在变得廉价了。但如何使用它、如何验证它，却变得更昂贵了。你愿意让你的图书管理员来当你的神经外科医生吗？>> [音乐] >> 你愿意让你的图书管理员来当空中交通管制员，去盯着几百架飞越北美或韩国上空的飞机吗？
04重新定义智力：创造与迁移
4:08
No way, because that human judgment is important. My identity has changed. My view on intelligence now has changed quite a bit. The ability to reason, make proper inferences, [music] whether you can do it quickly or slowly doesn't matter, but can you create a new concept? Can you generate ideas? >> [music] >> Can you string concepts together in a deep way? That is intelligence. That is not the regurgitation of facts. And we're not good at teaching that. Are you good at setting the dials >> [music] >> to design a system from scratch that was going to produce some gadget, whether it's as an industry [music] or a computer program or perhaps a whole new area of science. That's [music] deep intelligence. And it rarely is the form that is [music] recognized in schools at any level. Do you have the ability to recognize patterns [music] in areas of thought that can be transferred from one discipline to it another so that you can propel another area forward? I would have said 5 years ago, "Oh, that's just
绝对不行，因为人的判断力很重要。我的身份认同已经变了。我对智力的看法现在也变了很多。推理、做出恰当推断的能力，[音乐]无论你是快还是慢都不重要，但你能不能创造一个新概念？你能不能产生想法？>> [音乐] >> 你能不能把各种概念深刻地串联起来？那才是智力。那不是事实的复述。而且我们并不擅长教这个。你擅不擅长把各项参数调对 >> [音乐] >> 从零开始设计一个系统，它能造出某种装置，不管是作为一个产业[音乐]、一个计算机程序，还是也许一个全新的科学领域。那才是[音乐]深层的智力。而它很少是学校在任何层级上[音乐]所认可的那种形式。你有没有能力在思想的各个领域中识别模式，[音乐]并把它从一个学科迁移到另一个学科，从而推动另一个领域向前？五年前我会说：“哦，那只是在对的时间待在对的
5:16
being in the right spot at the right time." But that's unfair. Being in the right spot at the right time, but you still have to make that observation. So there's an element of recognizing a target of opportunity. That is genius, [music] and I don't use genius very lightly. The student, the worker who becomes an expert in their niche field [music] because they plug away and learn something new about that field every day is so hard-nosed and is [music] so committed that that is also intelligence and a kind of genius that we need to recognize.
地方。”但那样说不公平。在对的时间待在对的地方，但你仍然必须做出那个观察。所以这里有一个识别机遇目标的成分。那就是天才，[音乐]而我不会轻易使用“天才”这个词。那些成为自己细分领域专家的学生、工作者，[音乐]因为他们埋头苦干，每天都学到关于那个领域的新东西，如此坚韧，如此[音乐]投入，那也是一种智力，也是一种我们需要去认可的天才。
05数学家之子的叛逆少年
5:57
I have a very unique [music] personal story. That I'm a son of a mathematician. When I was a child, I was considered gifted in mathematics and my parents decided at an early age that I was going to be a mathematician. My parents [music] actually had a plan for all three of the boys. My oldest brother was going to be a pianist, he did it. I was the youngest [music] was going to be a mathematician and the middle son who they said wasn't good at math and wasn't gifted in music, well, he should just go work in a bank [music] just make a living which inspired him to do great things. My brother Santa uh has gone on to become the president of the University of Michigan. He's a [music] very distinguished scientist and he was literally driven by this need to prove that his early assessment was [music] incorrect. For me, it almost went in a very bad way. I dropped out of high school. The last thing I wanted to be in [music] high school was anything my parents wanted me to be. I didn't want to be the one Asian kid in class that
我有一个非常独特的[音乐]个人故事。我是一位数学家的儿子。我小时候被认为在数学上有天赋，我父母在很早的时候就决定我要成为一名数学家。我父母[音乐]其实对三个儿子都有规划。我大哥要成为钢琴家，他做到了。我是最小的，[音乐]要成为一名数学家；而他们说不擅长数学、也没有音乐天赋的二哥，嗯，他就该去银行工作[音乐]，混口饭吃就行，这反倒激励他做出了了不起的成就。我哥哥圣塔（Santa）后来成了密歇根大学的校长。他是一位[音乐]非常杰出的科学家，而他真的就是被这种想要证明早年那个评价[音乐]是错的的需求所驱动的。对我来说，事情差点就往很糟的方向发展了。我从高中辍学了。上[音乐]高中时我最不想成为的，就是我父母希望我成为的任何样子。我不想做班上那个被期待数学要好的亚裔孩子，而其他孩子都有自己的生活。我
06拉马努金遗孀的一封信
7:02
was expected to be good at math when all the other kids like had lives. I couldn't play baseball what you know, the all-American pastime. I I and I hated it. In April of 1984, I was of the mindset that I'm going to run away from home. I'm never going to see my parents again >> [music] >> and I don't care about that. That's I'm going to strike out on my own. In April 1984, a letter came to the house addressed to my father. It It was a on this yellowed piece of paper that looked like it might as well have been a hundred years old. It was a letter written by Janaki Ammal who was a widow of the Indian mathematician Ramanujan.
没法打棒球，你知道，那种全美国人的消遣。我，我讨厌那样。1984年4月，我当时的心态是：我要离家出走。我再也不见我父母了，>> [音乐] >> 我不在乎。就是，我要自己闯出一条路。1984年4月，家里来了一封信，收信人是我父亲。那是写在一张泛黄的纸上，看上去像是有一百年那么旧了。那是贾纳基·阿马尔（Janaki Ammal）写的信，她是印度数学家拉马努金的遗孀。
7:43
And she thanked my father for making a small gift to help commission a statue in in memory of Ramanujan. And me, seeing my dad cry, and he >> [music] >> never cried. He was very almost no emotions. If this letter brought him to tears, and he brought this letter to [music] me afterwards, said, "I I have to tell someone what this is about." So, Ramanujan, it turned out, I learned that day, was a mystic, an autodidact. He had visions of mathematics that his [music] goddess, he believed, gave him gifts of formulas that he would write down in his notebooks. Because of his passion for mathematics, he didn't study in any of his other courses, so he ended up flunking out of college twice. Here's my dad talking about someone who was a two-time college dropout, but had left behind three notebooks filled with formulas that he was studying himself.
她感谢我父亲捐了一笔小小的款项，帮忙促成一座纪念拉马努金的雕像。而我，看到我爸哭了，而他 >> [音乐] >> 从来不哭。他非常内敛，几乎没有情绪。如果这封信让他流泪了——之后他把这封信拿给[音乐]我看，说：“我，我得跟人说说这是怎么回事。”于是，那天我才知道，原来拉马努金是一位神秘主义者、一位自学成才的人。他有关于数学的异象，他相信是他的[音乐]女神赐予他公式作为礼物，他把这些写在自己的笔记本里。因为他对数学的痴迷，他别的课程都不学，结果两次从大学退学。这可是我爸爸在谈论一个两度大学辍学的人，但这个人留下了三本写满公式的笔记本，是他自己研究出来的。
07父亲的战后人生与希望
8:40
And I only learned later one of the reasons that my father was so in love [music] with the story of Ramanujan is because Ramanujan had actually represented hope for him, his one chance in life as a Japanese mathematician post-World War II. My father wanted to be a mathematician, but he went to college at a time when the world was at war. He loved mathematics as a way of escaping from [music] long lines waiting for food. In the aftermath of World War II, the United States sent some of their [music] best mathematicians to Japan to rebuild the universities and train the mathematicians. My father first learned about Ramanujan at the [music] conference where he was discovered by a Princeton professor who invited him to study [music] with him at Princeton, launching his career.
而我后来才知道，我父亲之所以如此钟情[音乐]于拉马努金的故事，一个原因是拉马努金对他而言代表着希望，是他作为二战后一名日本数学家此生唯一的机会。我父亲想成为数学家，但他上大学的时候正逢世界大战。他热爱数学，把它当作逃离[音乐]排长队领食物的一种方式。二战之后，美国派了一些他们[音乐]最优秀的数学家到日本，去重建大学、培养数学家。我父亲第一次听说拉马努金，是在那场[音乐]会议上，他在那里被一位普林斯顿的教授发现，那位教授邀请他到普林斯顿[音乐]跟随自己学习，从而开启了他的事业。
9:34
Ramanujan died a very early age and at [music] 32. He was nearly forgotten and the mathematicians of the world contributed small gifts to give Mrs. Ramanujan the statue that the government had promised her in 1920. That letter and a photograph that she shared with us of the statue, it represented him remembering what it was like for him [music] to struggle and the moment where he got his chance. So, what did Ramanujan mean to me that day? It gave me hope in the following. [music] It was the first time I heard my parents say and look up to like a hero someone who hadn't gone to Harvard or Princeton and was a perfect student. On the contrary, it turned out my father's hero was a two-time college dropout.
拉马努金在很年轻的时候就去世了，[音乐]年仅32岁。他几乎被遗忘了，世界各地的数学家们捐了一些小额款项，为拉马努金夫人立起了政府在1920年答应过她的那座雕像。那封信，还有她与我们分享的雕像照片，代表着他回想起自己曾经[音乐]如何挣扎，以及他获得机会的那一刻。那么那天拉马努金对我意味着什么呢？它在以下这一点上给了我希望。[音乐]那是我第一次听到我父母提起、并像敬仰英雄一样敬仰一个没上过哈佛或普林斯顿、也不是完美学生的人。恰恰相反，原来我父亲的英雄是个两次辍学的大学生。
08追随拉马努金的学术转折
10:23
And I needed [music] that. Later at the University of Chicago, I was a horrible student. But right before my senior year, flipping through the channels on the television, I saw a video, Public Broadcasting Service documentary about Ramanujan who I hadn't thought about in years and I was fascinated because here on in color on TV were was more than the vaguest of outlines about Ramanujan that my dad told me. There was the whole story and it kind of jump-started me. I had a lot of catching up to do. There I became a good student and it was then [music] when the biography of Ramanujan came out called The Man Who Knew Infinity. Maybe it was a sign. Maybe I was meant to follow Ramanujan and so I started to work on a thesis [music] based on his work and I'm so glad I made that choice because by the end of my PhD uh what I worked [music] on was called the theory of Galois representations which was meant to study Ramanujan's backwater mathematics.
而我需要[音乐]那个。后来在芝加哥大学，我是个糟糕的学生。但就在大四前不久，我在电视上换台时，看到了一部公共广播公司（PBS）的纪录片，讲的是拉马努金——一个我好多年没想起过的人——我被迷住了，因为电视上彩色画面里呈现的，远不止我爸告诉我的那些关于拉马努金的最模糊的轮廓。那里有完整的故事，它某种程度上把我点燃了。我有很多要补的功课。在那之后我成了一个好学生，也正是在那时[音乐]拉马努金的传记出版了，书名叫《知无涯者》（The Man Who Knew Infinity）。也许那是个征兆。也许我注定要追随拉马努金，于是我开始做一篇[音乐]基于他工作的论文，我非常庆幸自己做了那个选择，因为到我博士读完时，我研究的[音乐]东西叫做伽罗瓦表示论，它本来是用来研究拉马努金那些冷门数学的。
09寻找散落人间的拉马努金
11:24
But by 1993, the bombshell news in mathematics for the end of the 20th century was a proof of Fermat's Last Theorem And the proof of Fermat's Last Theorem depend on these Galois representations. I don't know what it is. Following Ramanujan every time he appeared has been like the best decision I've ever made in my life. And every one of those could have gone a different way. So, one important theme about Ramanujan for me is where would we be? And I don't mean just me as a mathematician. [music] Where would we all collectively be had he not been discovered? That is a world I cannot fathom. And because of that, you're left wondering there must be other Ramanujans walking planet Earth. Maybe they don't come from privilege. How do we find them?
但到了1993年，20世纪末数学界的爆炸性新闻是费马大定理的证明，而费马大定理的证明依赖于这些伽罗瓦表示。我不知道这算什么。每一次拉马努金出现时都去追随他，一直是我这辈子做过的最好的决定。而其中每一次都可能走向完全不同的结局。所以对我来说，关于拉马努金的一个重要主题是：我们会在哪里？我说的不只是我作为一名数学家。[音乐]如果他没有被发现，我们所有人集体会在哪里？那是一个我无法想象的世界。正因如此，你不禁会想，一定还有别的拉马努金行走在地球上。也许他们并非出身优渥。我们怎样才能找到他们？
12:14
And how do we nurture them when we find them? I was lucky enough for a number of years to run a program called the Spirit of Ramanujan, where we looked for undiscovered [music] talent. And what's interesting about this, my boss, my former student Karina Hong, studied with me in our a research program. She was one of our first recipients. We discovered her. It makes me wonder where she would be today had she not [music] received this Spirit of Ramanujan fellowship. There are, I am sure, many, many undiscovered folks that we need to find. I think the idea, the ability and the potential [music] to be someone like Ramanujan, or at least creative in a productive way, I think it resides in us all. You just have [music] to give students of all ages the opportunity to A be brave enough to act on their curiosity, >> [music] >> and then offer them a system that embraces it.
找到之后又怎样去培养他们？我有幸在好些年里主持一个叫“拉马努金精神”的项目，我们在其中寻找尚未被发掘的[音乐]人才。而有意思的是，我的老板、我以前的学生卡里娜·洪（Karina Hong），曾在我们的研究项目里跟我学习。她是我们最早的受助者之一。是我们发掘了她。这让我不禁想，如果她当年没有[音乐]得到这份“拉马努金精神”奖学金，她今天会在哪里。我确信，还有非常非常多未被发掘的人，是我们需要去找到的。我认为那种成为像拉马努金那样的人、或者至少能以富有成效的方式进行创造的想法、能力和潜力，[音乐]我认为它存在于我们每个人身上。你只需要[音乐]给各个年龄段的学生机会：第一，勇敢到敢于按自己的好奇心去行动，>> [音乐] >> 然后给他们提供一个能接纳这种好奇心的体系。
10打勾式教育的批判
13:13
Some of your best students in Korea, the best students in the United States, best students worldwide, they're stressed out in high school. They're probably even stressed out in in middle school worrying about how do I get into the right high school? How do I get into the right college? Will I get the right test scores? If you're motivated to participate in those just because they are checkboxes, that's messed up. And I'm not saying that you shouldn't do that because I don't want to be ignorant and say don't participate in the system that will ultimately decide your fate with regard to college, but pause and recognize you're participating in that system.
你们韩国最优秀的一些学生、美国最优秀的学生、全世界最优秀的学生，他们在高中时压力巨大。他们大概在初中就已经压力巨大了，担心我怎样才能进对的高中？我怎样才能进对的大学？我能不能考出对的分数？如果你参与这些事情的动力只是因为它们是一个个要打勾的方框，那就不对劲了。而我并不是说你不该那么做，因为我不想无知地说：别去参与那个最终会决定你大学去向的体系；但请停下来，意识到你正在参与那个体系。
13:48
Education starts with inspiring people to want to know more about the world in which they live, wanted to know more about the cultures of the world because we share the world together, and appreciate what is different in other parts of the world and in other cultures because that's why I went to college, right? I wanted to learn about those things. That's why I travel the world. If [music] you have children that are infants, how wonderful is it to play with them, say with like a stack of boxes or building blocks. Play for children is science. They're not really learning about gravity, but they're really learning about gravity. They may pile blocks on top of the other and knock them over and giggle, and they'll do it again. Think about how wonderful the world is >> [music] >> when you get to learn about it without worrying about what your future and what your reputation will be. [music] I want when my students are in my class to say this is a wonderful class, Professor Ono, because the subject is
教育始于激励人们想要更多地了解他们所生活的这个世界，想要更多地了解世界各地的文化，因为我们共享这个世界，并且欣赏世界其他地方、其他文化中不同的东西，因为那正是我上大学的原因，对吧？我想学习那些东西。那也是我周游世界的原因。如果[音乐]你有还是婴儿的孩子，跟他们一起玩是多么美妙的事，比如玩一摞盒子或积木。对孩子来说，玩耍就是科学。他们并不是真的在学重力，但他们真的是在学重力。他们可能把积木一块块垒起来，然后推倒，咯咯地笑，然后再来一次。想想这个世界有多么美妙 >> [音乐] >> 当你可以去认识它，而不必担心你的未来、你的名声会怎样。[音乐]我希望当我的学生在我课堂上时，他们会说：这是一门很棒的课，小野教授，因为这门学科很美。
14:50
beautiful. And I do my very best to try to get that across, but I'm not a fool. I know that I'm participating in a system where [music] at the end of the day the students are worried, am I going to get an A or not, [music] and how will that impact my ability to go to graduate school in math or medical school or law school because that GPA is so [music] important. And I hate that. I utterly hate that. Why? It's an opportunity lost. What I like about AI, and this is actually how I transitioned from being devastated by AI's already read my papers. It understands my papers better than I remember them.
而我会尽我所能把这一点传达出去，但我不傻。我知道我身处的这个体系里，[音乐]说到底，学生们担心的是：我能不能拿到A，[音乐]以及这会怎样影响我去读数学研究生、医学院或法学院的机会，因为绩点太[音乐]重要了。我讨厌这一点。我极其讨厌这一点。为什么？因为这是一种机会的丧失。我喜欢AI的地方——其实这正是我如何从“AI已经读过我的论文”的打击中走出来的。它对我论文的理解比我自己记得的还要好。
11知识廉价后大学剩下什么
15:31
>> [music] >> Think of all the subjects in adjacent areas of mathematics that I could just ask AI about, and AI is not going to laugh at me, and it as a great librarian will dutifully answer [music] any question I would ask. The access to knowledge, if you are privileged enough to have access to the internet, and [music] you are privileged enough to be able to afford access to a large language model, knowledge quickly became cheap. In the United States, it could cost $80,000 to spend 1 year attending a university.
>> [音乐] >> 想想数学相邻领域里所有那些学科，我都可以直接去问AI，而AI不会嘲笑我，它像一位出色的图书管理员，会尽职地回答[音乐]我提出的任何问题。获取知识这件事，如果你足够幸运能上网，[音乐]并且足够幸运能负担得起使用大语言模型，知识很快就变得廉价了。在美国，上一年大学可能要花8万美元。
16:03
>> [music] >> And here's the dirty secret. I could learn everything that you would learn book-wise, academically, from a large language model at my own pace, probably accelerated >> [music] >> with a large language model. What I would not get would be the human access, how the right questions were derived, [music] what the next questions in a field might be. That's why we still go to college, and that's why we still need professors. But all of the other stuff, the tutoring, the precision learning, that AI can help with. I actually believe that we in this world aren't doing the best we can at educating our children. And I don't say that to be critical of educators. I am an educator.
>> [音乐] >> 而这里有个不能说的秘密。我可以从大语言模型那里学到你在书本上、在学业上会学到的一切，按我自己的节奏，>> [音乐] >> 有了大语言模型可能还会更快。我得不到的是人与人之间的接触，那些正确的问题是怎么被提出来的，[音乐]一个领域接下来的问题可能是什么。这就是我们仍然要上大学的原因，也是我们仍然需要教授的原因。但其他所有那些东西，辅导、精准化学习，AI是可以帮上忙的。我其实相信，在教育我们的孩子这件事上，我们这个世界还没有做到最好。我这么说并不是要批评教育工作者。我自己就是教育工作者。
12守护惊奇感与身份所有权
16:50
But it's always a treat to visit a kindergarten class, a first grade class, when it's bring your parent to school day so they can talk about what they do. "Oh, I know all the prime numbers." Or "I'm really good at adding." That wonder. And I want to just bottle up this energy. Because if we could maintain that wonder in the world and the energy that children have when everything around them is new, think about where we would be today. They go find your passion. The best scientists in the world need to still view the world as a wondrous thing.
但去参观幼儿园的课堂、一年级的课堂，总是一种享受，尤其是“带家长来学校日”，家长可以讲讲自己是做什么的。“哦，我知道所有的质数。”或者“我特别会做加法。”那种惊奇感。我真想把这股能量装进瓶子里。因为如果我们能在这个世界上保住那种惊奇感，保住孩子们在周遭一切都是新鲜事物时所拥有的那种能量，想想我们今天会走到哪一步。他们说去找到你的热情所在。世界上最好的科学家仍然需要把这个世界看作一件奇妙的事物。
17:28
>> [music] >> The best doctors in the world still need to recognize that what they practice is supposed to come from a place of benevolence, not I have this practice, but I'm a university professor that happens to be has a clinical practice and I'm going to write articles about my patients. I think that's messed up. If we pay so much attention on and value so much perfection and speed in ordinary test taking, >> [music] >> then how are we training someone to be the next Einstein or the next you name [music] famous professor was just wondering out loud in their lab, I wonder if such and such is true. For my children, I want them to be passionate about the world that they live in. And if you're passionate about the world that you're you're living, then you're deeply worried about the climate. You're deeply worried about the conflict [music] that exists between different cultures, which this war is all over the place. How can that happen?
>> [音乐] >> 世界上最好的医生仍然需要认识到，他们所从事的行医，本应出自一种仁爱之心，而不是“我有这么个诊所，但我是个大学教授，恰好还有临床门诊，我要拿我的病人写文章”。我觉得那样不对劲。如果我们把那么多注意力和价值放在……很讲究普通考试中的完美和速度，>> [音乐] >> 那我们又怎么去培养下一个爱因斯坦，或者下一个随便你说得出名字的人[音乐] 有名的教授只是在实验室里自言自语，我在想某某事情是不是真的呢。对我的孩子，我希望他们能对自己生活的这个世界充满热情对这个世界满怀热情。如果你对自己所生活的世界充满热情，那么你就会深深地为气候担忧。你会深深地为不同文化之间存在的冲突而担忧 [音乐]这种冲突到处都是，这场战争到处都是。这种事怎么会发生？
18:26
Who are the people that you really look up to the most? Maybe they are the odd balls. In [snorts] in my country, the United States, where education is so expensive, you could come out of college with $150,000 in debt. And then you may go on to a professional school accruing another $200,000 in debt. And [music] then 3 years later discovering, I can't stand the sight of blood, but now I can't leave my profession because I have all of these loans. That is purgatory. You then are stuck. And that's when your life is, well, I go to work >> [music] >> because it pays the bills. Who owns your identity? You do.
你最敬佩的人都是谁？也许他们是那些怪人。在 [吸气声] 在我的国家美国，教育非常昂贵，你大学毕业时可能就背着 15 万美元的债务。然后你可能还要去读专业学院，再背上 20 万美元的债务。然后 [音乐]三年后你发现，我受不了看见血，但现在我又没法离开这个职业，因为我背着这么多贷款。那简直是炼狱。你就被困住了。到那时你的人生就变成了，唉，我去上班 >> [音乐] >> 是因为它能付账单。谁拥有你的身份？是你自己。
19:30
>> Mhm.
>> 嗯。
视频总结 · 一句话概括与核心要点
一句话概括
数学家 Ken Ono 讲述自己在 Frontier Math 项目中被大语言模型"击溃"后，如何重新定义智能与教育的价值：知识已经廉价，真正稀缺的是判断、创造概念与发现"下一个 Ramanujan"的能力。
核心要点
- "跑赢 AI"是错误的问题：一年前 Ono 受 Epoch AI 之邀为 Frontier Math 出题评测模型，结果第一次发现自己很难出出 ChatGPT 答不对的题，为此沮丧了数月；他最终的结论是，若目标是永远领先 AI，人类必输——正如没人想看博尔特和摩托车赛跑，但奥运会依然存在。
- LLM 是史上最强的图书馆员，而非决策者：凡是被写下、被上传到 YouTube、甚至当天早上刊登的报纸内容，模型几乎都已"读过"；但你不会让图书馆员做神经外科手术或指挥北美上空数百架飞机——知识变便宜了，如何使用和验证知识变贵了。
- 智能的定义被迫改写：不是复述事实，而是创造新概念、把概念深度串联、跨学科迁移模式、从零设计系统；他同时强调"每天钻研一点、成为冷门领域专家"的执着也是一种天才，学校体系几乎不识别这两类能力。
- Ramanujan 三次改写他的人生：1984 年 Ramanujan 遗孀的一封感谢信让从不流露情感的父亲落泪，让正打算离家出走的高中辍学生 Ono 第一次听到父母把一个"两度从大学退学的人"当英雄；大学末期看到 PBS 纪录片重新振作；博士论文选了 Ramanujan 相关的 Galois 表示理论，恰好是 1993 年费马大定理证明的核心工具。
- 父亲的执念来自战后日本：父亲在二战食物配给的长队中靠数学逃避现实，战后美国派数学家赴日重建大学，他在会议上被普林斯顿教授发掘——Ramanujan 对他而言就是"被发现"的希望。Ono 的哥哥因被父母断言"不适合数学"而发奋，后来成为密歇根大学校长。
- "还有多少 Ramanujan 未被发现"是核心公共议题：Ono 曾主持 Spirit of Ramanujan 项目寻找未被发掘的人才，他现在的老板、Axiom Math 的 Karina Hong 正是首批受资助者之一；他相信创造潜能人人皆有，缺的是敢于追随好奇心的勇气和接纳它的体系。
- AI 暴露了大学的"肮脏秘密"：美国一年学费可达 8 万美元，而书本知识完全可以靠 LLM 自学甚至加速；大学和教授剩余的真正价值在于人际接触、学会如何提出正确问题、看到领域的下一个问题——辅导和精准学习则应交给 AI。
- 考试型完美主义在扼杀好奇心：韩国、美国最优秀的学生从初中起就为择校和分数焦虑，把学习当作打勾；他呼吁"瓶装"幼儿园孩子玩积木时的那种好奇（玩耍即科学），因为顶尖科学家必须仍把世界当作奇迹，而不是靠速度和准确率训练出来的答题机器。
- 高额学费债务造成"炼狱"：15 万美元本科债务加 20 万专业学院债务后，若发现自己"见血就晕"也无法转行，人生沦为"上班只为还账单"——他的收尾问题是：谁拥有你的身份？只有你自己。
结论与值得注意的细节
- Ono 的立场转变清晰：从"被 AI 击溃"到"把 AI 当作不会嘲笑你的全知图书馆员"，用它跨到相邻数学领域学习，而把人类精力集中在判断、验证与提出新问题上。
- 他强调自己是内部人（教育者、正在参与 GPA 体系的教授），批评的是体系而非教师——承认学生"不得不"参与升学竞争，但要求他们至少意识到自己在参与什么。
- 细节：他 1993 年初识"人工智能"一词时自称"专攻自然智能"，如今自嘲那句狂言；Ramanujan 32 岁早逝，1920 年政府承诺的雕像直到 1980 年代才靠全球数学家的小额捐款落成。
- 演讲从 AI 开始，结尾却落在教育与身份：在知识免费的时代，教育的起点应当是激发对世界和不同文化的好奇，而非制造完美的考生。
核心句型 · 9
1. I struggled to do … that … would get wrong
“For the first time, I struggled to assemble questions that ChatGPT would get wrong.”
struggle to + 动词表「费尽力气仍近乎做不到」，后接定语从句限定对象。表达「连……都难以做到」的挫败感时，比 could hardly 更有画面感，可仿写工作汇报中的瓶颈描述。
2. If our goal is to …, then … we're going to lose
“If our goal is to always stay ahead of AI, then I think we're going to lose.”
用条件句把对方的目标本身摆上台面再宣判结局，实现「问题重构」（reframing）。辩论或演讲中否定一个流行前提时极好用：先复述其目标，再推出必败结论。
3. Nobody would be interested in watching … But we still …
“Nobody would be interested in watching Usain Bolt race against a motorcycle in the 1-mile run. … But we still watch the Olympics.”
先立一个具体到荒谬的假想画面，再用 But we still 转折，把类比迁移到真正的论题。说理时用具象场景替代抽象概念，是英语演讲的经典开局手法。
4. X should be thought of as …
“The large language models should be thought of as the most extraordinary librarian the world has ever seen.”
被动式「应被看作……」用于给争议对象重新下定义，比 X is … 委婉且更具提议性；后接 the most … the world has ever seen 最高级加码，适合观点输出型写作。
5. If it has been …, X has probably …
“If it has been written down, the large language model has probably seen it. If it's on YouTube, large language model has probably been trained on it.”
排比条件句 + probably 弱断言：单句留有余地，连续堆叠却形成压倒性证据感。列举某事物无孔不入的覆盖面时可直接套用这一「排比+概率词」结构。
6. Do you want your X to be your Y?
“Do you want your librarian to be your neurosurgeon?”
反问句连发划定能力边界：X 与 Y 错位越大，荒谬感越强，答案不言自明。反驳「A 很强所以能替代 B」类观点时，是比正面论证更省力的修辞。
7. The last thing I wanted to be … was …
“The last thing I wanted to be in high school was anything my parents wanted me to be.”
the last thing 表最强烈的否定意愿，比 I really didn't want 语气重且地道。本句还嵌套 anything my parents wanted me to be，双层从句值得整句背诵。
8. Where would we be had he not been …?
“Where would we all collectively be had he not been discovered?”
虚拟条件句省略 if 的倒装形式：had + 主语 + 过去分词，正式书面语色彩浓。表达「若无此人/此事，历史将如何」的反事实设问，演讲收尾常见。
9. What I would not get would be …
“What I would not get would be the human access, how the right questions were derived.”
what-伪分裂句把「缺失的东西」提到句首聚焦，双重 would 保持虚拟语气一致。对比「能得到什么/得不到什么」时，这一结构比 but I can't get 更有强调力。
生词精讲 · 57 · 按出现顺序
happy-go-lucky
/ˌhæpi ɡoʊ ˈlʌki/
adj.
0:00
无忧无虑的，随遇而安的
devastated
/ˈdevəsteɪtɪd/
adj.
0:00
深受打击的，崩溃的（比 upset 程度重得多）
stay ahead of
phr.
0:00
保持领先于……
cocky
/ˈkɑːki/
adj.
1:07
自大的，过分自信的（口语，略带贬义）
have my butt handed to me
phr.
1:07
被狠狠教训、彻底击败（俚语，常用被动式自嘲）
came face-to-face with
phr.
1:07
与……正面遭遇，直面
state-of-the-art
/ˌsteɪt əv ði ˈɑːrt/
adj.
2:17
最先进的，顶尖水平的
outperform
/ˌaʊtpərˈfɔːrm/
v.
2:56
表现优于，胜过
coming to grips with
phr.
2:56
开始正视并设法应对（棘手的现实）
inquiry
/ˈɪnkwəri/
n.
2:56
探究，钻研（deep inquiry 深度思考）
neurosurgeon
/ˌnʊroʊˈsɜːrdʒən/
n.
3:37
神经外科医生
regurgitation
/rɪˌɡɜːrdʒəˈteɪʃən/
n.
4:08
机械复述；原义为反刍，喻不加消化地照搬知识
inferences
/ˈɪnfərənsɪz/
n.
4:08
推断，推论（make inferences 固定搭配）
string concepts together
phr.
4:08
把概念串联起来（string...together 把零散事物连成整体）
from scratch
phr.
4:08
从零开始，白手起家
gadget
/ˈɡædʒɪt/
n.
4:08
小装置，小发明
propel
/prəˈpel/
v.
4:08
推进，推动（某领域）向前
setting the dials
phr.
4:08
调准各项参数（喻：为系统设定正确条件）
target of opportunity
phr.
5:16
临机目标；意外出现、须当即抓住的机会（源自军事用语）
hard-nosed
/ˌhɑːrdˈnoʊzd/
adj.
5:16
顽强务实的，不屈不挠的
plug away
phr. v.
5:16
埋头苦干，坚持不懈地做（常接 at sth）
niche
/nɪtʃ/
adj./n.
5:16
细分的、小众的（领域）
distinguished
/dɪˈstɪŋɡwɪʃt/
adj.
5:57
卓越的，声望崇高的（常修饰学者、生涯）
strike out on my own
phr.
7:02
独立闯荡，自立门户
pastime
/ˈpæstaɪm/
n.
7:02
消遣，娱乐活动（all-American pastime 指棒球）
yellowed
/ˈjeloʊd/
adj.
7:02
（纸张因年久）泛黄的
commission
/kəˈmɪʃən/
v.
7:43
委托制作（艺术品、作品）
mystic
/ˈmɪstɪk/
n.
7:43
神秘主义者
autodidact
/ˌɔːtoʊˈdaɪdækt/
n.
7:43
自学成才者
flunking out
phr. v.
7:43
因成绩不及格而退学（flunk out of college）
aftermath
/ˈæftərmæθ/
n.
8:40
（战争、灾难后的）余波时期，善后阶段
look up to
phr. v.
9:34
敬仰，视为榜样
jump-started
/ˈdʒʌmpstɑːrtɪd/
v.
10:23
（如搭电启动汽车般）激活，重新点燃
backwater
/ˈbækwɔːtər/
n.
10:23
死水；喻冷门、无人问津的领域
bombshell
/ˈbɑːmʃel/
n.
11:24
爆炸性消息，令人震惊的事件
fathom
/ˈfæðəm/
v.
11:24
彻底理解，想象（多用于否定：cannot fathom）
come from privilege
phr.
11:24
出身优渥，来自特权阶层
nurture
/ˈnɜːrtʃər/
v.
12:14
培养，养育（人才、能力）
fellowship
/ˈfeloʊʃɪp/
n.
12:14
（资助学者/学生的）奖学金、研究基金
recipients
/rɪˈsɪpiənts/
n.
12:14
获得者，受助人
resides
/rɪˈzaɪdz/
v.
12:14
存在于，蕴藏于（reside in sb/sth，正式用语）
messed up
phr./adj.
13:13
不对劲的，糟糕透顶的（口语，表道德或状态上的错乱）
giggle
/ˈɡɪɡəl/
v.
13:48
咯咯地笑
get that across
phr. v.
14:50
把（观点、感受）传达给别人（get sth across to sb）
utterly
/ˈʌtərli/
adv.
14:50
完全地，彻底地（加强语气，多修饰负面词）
dutifully
/ˈduːtɪfəli/
adv.
15:31
尽职地，恭顺地
adjacent
/əˈdʒeɪsənt/
adj.
15:31
相邻的，毗连的（adjacent areas 相邻领域）
dirty secret
phr.
16:03
不可告人的秘密；行业内心照不宣的真相
a treat
/triːt/
n.
16:50
乐事，难得的享受（it's a treat to do sth）
bottle up
phr. v.
16:50
封存起来；此处指想把能量「装进瓶子里」保存（另义：压抑情绪）
wondrous
/ˈwʌndrəs/
adj.
16:50
奇妙的，令人惊叹的（文学化用词）
benevolence
/bəˈnevələns/
n.
17:28
仁爱，善意
wondering out loud
phr.
17:28
自言自语地琢磨，把心中疑问说出声（wonder out loud）
odd balls
/ˈɑːdbɔːlz/
n.
18:26
怪人，与众不同的人（常拼作 oddballs，略带亲昵）
accruing
/əˈkruːɪŋ/
v.
18:26
（债务、利息等）逐渐累积
purgatory
/ˈpɜːrɡətɔːri/
n.
18:26
炼狱；喻进退两难的煎熬处境
pays the bills
phr.
18:26
挣钱糊口（暗示工作仅为谋生、无热情）
理解自测 · 11 题 · 是真懂了，还是以为自己懂
1. 一年前让 Ono 从「无忧无虑的教授」变得「深感不安」的具体事件是什么？
他参与了伯克利公司 Epoch AI 发起的 FrontierMath 项目：该项目雇佣全球职业数学家出极难的数学题，用于评测大语言模型不断迭代的能力。作为出题人和少数拿到最先进模型内测权限的科学家，他第一次发现自己「很难出出能让 ChatGPT 答错的题」，并且要等几个月世界才承认这些模型知道的事实超过任何人。这段经历让他崩溃了数月（见开头「FrontierMath 与崩溃时刻」两章）。
2. 他用什么比喻来定位大语言模型的能力？这个比喻的边界在哪里？
他称 LLM 为「世界上有史以来最了不起的图书馆员」：凡被写下的内容它大概率见过，YouTube 视频、当天上午的报纸文章都可能已进入它的视野，人类在信息收集上无法与之竞争。但边界同样清晰——你不会让图书馆员做神经外科手术，也不会让它指挥几百架飞越北美或韩国上空的飞机，因为这些岗位要的是人的判断与责任，不是知识检索（「史上最强图书馆员」章）。
3. 1984 年 4 月寄到 Ono 家的那封信是谁写的？为什么让他父亲落泪？
信出自拉马努金的遗孀贾纳基·阿马尔，写在一张泛黄如百年旧物的纸上，感谢 Ono 的父亲捐款帮助铸造拉马努金纪念雕像——印度政府 1920 年就承诺立像却拖延了 60 多年，最终靠全球数学家的小额捐款实现。几乎从不流露情绪的父亲被这封信感动落泪，并第一次向儿子讲述拉马努金：一个两度辍学、自学成才、自称公式来自女神启示的天才（「拉马努金遗孀的一封信」章）。
4. Ono 的博士研究方向与费马大定理之间有什么戏剧性的联系？
大四前他偶然看到 PBS 关于拉马努金的纪录片而被点燃，从芝加哥大学的差生变成好学生，随后传记《知无涯者》出版，他索性以拉马努金的工作为论文方向，研究当时被视为「冷门数学」工具的伽罗瓦表示论。1993 年怀尔斯宣布证明费马大定理——20 世纪末数学界的爆炸性新闻——证明恰恰依赖伽罗瓦表示。他由此感叹：每次追随拉马努金，都是一生中最好的决定（「学术转折」章）。
5. 博尔特与摩托车赛跑的类比，在他的论证中起什么作用？
类比承担两步推理：第一，没人想看博尔特和摩托车比一英里赛跑，因为那不是公平比赛；第二，机器在一切体能上超越人类之后，我们照样看奥运会——说明社会早已学会把「机器更强」与「人的价值」分开。他要求对脑力做同样的切割：既然 LLM 在事实性知识上必胜，「如何领先 AI」就是选错了赛道，注定要输；值得问的是人的判断、创造与验证能力该用在哪条赛道上（「图书馆员」与开场章）。
6. 在 Ono 重新定义的「智力」里，包含哪些成分？为什么他说学校不教这些？
他列出的成分包括：做出恰当推断（快慢无关）、创造新概念、把概念深度串联、从零「调对参数」设计出新系统（产业、程序或新学科）、识别可跨学科迁移的思维模式；此外还有两种被低估的天才——抓住临机出现的机遇目标的洞察力，以及在细分领域每天精进的坚韧。这些的共同点是「生成新东西」，而学校各层级认可的主要是事实复述与考试的完美和速度，所以他说「我们不擅长教这个」（「重新定义智力」章）。
7. 拉马努金的故事为何恰好在那个时点「救了」少年 Ono？这与他父亲的经历构成怎样的镜像？
1984 年 4 月他正处于要离家出走、与父母期望决裂的心态。信到来后，他第一次听到父母像敬仰英雄一样谈论一个「没上过哈佛普林斯顿、两度辍学」的人——这证明不走既定完美路径也能有价值，他说「我需要那个」。对父亲则是镜像：战后日本物资匮乏，数学是逃离排队领食物的出口，拉马努金代表他「此生唯一的机会」，而他本人正是在国际会议上被普林斯顿教授发现才开启学术生涯。同一个故事在两代人身上完成了同样的救赎。
8. 从「知识变廉价」出发，他如何推导出大学与教授的剩余价值？
推导链是：LLM 几乎读过一切书面知识、按个人节奏尽职回答且不会嘲笑提问者，因此书本层面的学习可以被极低成本替代，一年 8 万美元的大学若只卖「知识传递」已不划算——这是他所说的 dirty secret。但 LLM 给不了三样东西：人与人的接触、「正确的问题当初如何被提出」、「这个领域下一个问题是什么」。于是大学与教授的核心价值收缩并聚焦到提问传统与判断力的传承，而辅导、精准化学习应交给 AI（「知识廉价后大学剩下什么」章）。
9. 他对应试体系的态度为何是「参与但保持自觉」，而不是号召退出？
他明确说不想「无知地」劝学生退出这个最终决定大学去向的体系——那对个体不负责任；但如果参与的唯一动机是把一个个方框打上勾，那就是 messed up。他的方案是双重意识：一边参与，一边清醒地知道自己在参与什么，同时守住好奇心与惊奇感这一真正的学习动力。他也承认自己身在其中：他极其痛恨学生只关心 GPA，却明白自己就是这个体系的参与者——批判与现实感并存（「打勾式教育的批判」章）。
10. 如果有人反驳：「AI 终将学会创造概念与跨学科迁移，你的智力新定义不过是又一次退守」，按片中逻辑 Ono 会如何回应？
可以从他的论证里提取两层回应。其一，他的核心主张本就不是「AI 永远做不到」，而是「人的价值不该建立在对机器的绝对优势上」——正如体能被机器全面超越后奥运会依然成立，即使 AI 逼近创造力，人类活动的意义也不因此清零。其二，他反复强调判断伴随责任：神经外科、空管的例子说明社会对「由谁担责」的要求不会随 AI 能力提升而消失，「如何用、如何验证、谁负责」仍是人的问题。当然，这一回应能否长期成立，片中并未给出证明，属于他的信念性判断。
11. 「寻找散落人间的拉马努金」这一使命，放到 AI 普及的时代还成立吗？
按片中逻辑，不仅成立而且被放大。拉马努金当年靠一封寄往剑桥的信才被发现，偶然性极高；Ono 之问「如果他没被发现，我们所有人会在哪里」指出瓶颈正在发现与培养的渠道。若 LLM 让任何能上网的人以极低成本接触前沿知识、获得不带嘲笑的辅导，出身造成的知识壁垒被削弱，未被发现的天才更容易自己冒头——他的 Spirit of Ramanujan 项目和受助者 Carina Hong 后来创立 Axiom 就是例证。但他也点出前提是双重 privilege（能上网、付得起模型），且仍需一个「接纳好奇心的体系」，否则发现了也养不活。
下载便签
手机：长按图片保存
← 上一期 · REC_015MIT Godel Escher Bach Lecture 2
下一期 · REC_017 →Demis Hassabis on AI's Next Big Breakthrough, 2050 and More!
苏菲拉底
THE SOPHIE LAB · ASK THE BEST MINDS THE BIG QUESTIONS
内容仅供学习 · thesophielab.com