李录：哥大商学院第36届格雷厄姆与多德年度论坛对谈实录（2026-10-05）

基本信息

• 活动：哥伦比亚大学商学院第36届格雷厄姆与多德年度论坛（36th Annual Graham & Dodd Luncheon），海尔布伦格雷厄姆与多德投资中心主办
• 时间：2026 年 10 月 5 日 12:00–13:30（纽约时间）
• 地点：纽约大都会俱乐部（Metropolitan Club，1 East 60th Street）
• 主讲：李录（喜马拉雅资本创始人及董事长，哥大 1996 届校友）
• 形式：主持人对谈加现场提问。主持人为哥大商学院教授 Tano Santos（李录在对谈中称其为 Tano）
• 语言：英文

来源与整理说明

截至 2026 年 10 月 7 日，哥大海尔布伦中心和喜马拉雅资本官网都没有发布这场活动的文字稿或录像。海尔布伦中心历年都不公开这场午餐会的录像，官方页面从第 1 届到第 35 届均无回放链接。

本文依据东方港湾董事长但斌在现场拍摄并发布的视频整理：

• 来源：但斌微博，2026 年 10 月 6 日 05:33（北京时间）发布，正文为"哥大商学院第36届格雷厄姆与多德年度论坛（Graham and Dodd Event），一场价值投资的殿堂级盛会。这是传承格雷厄姆价值投资思想的顶级活动，今年主讲人为李录先生……"。链接：https://weibo.com/1249424622/5350894137770030 。同一批视频也发在他的微信视频号。
• 视频：该微博附 8 段视频，其中 4 段为长片段，合计有效时长约 44 分钟（活动全长 90 分钟）。第 7 段与第 5 段内容重复（同一录像的两次上传），已剔除。
• 转写：用 whisper large-v3-turbo 模型本地转写英文原声，再人工按语义顺序拼接、翻译为中文。转写中明显的听写错误已按上下文订正（例如"two versus creature"订正为"tool versus creature"、"the bircher"订正为"Berkshire"、"Mexico plus Xi'an"订正为"Mexico plus Vietnam"、"Li Pu"订正为"Lip-Bu"）。听不清或无法确认的地方在文中标注。
• 覆盖范围：片段之间有缺口。六条原则中前两条的大部分、提问环节的部分问题没有拍到。本文不是官方全文，官方文字稿待喜马拉雅资本发布后以官方版本为准。

片段顺序与衔接

对谈中文全文

一、主持人开场（片段 2）

主持人：我们为学生组合，也就是汤姆·鲁索学生组合，做了新的选股。今年我们买了 Equifax、Humana，还买了伟大的日本自行车零部件公司禧玛诺。

（此后一段未拍到。）

二、价值投资的六条基本理念（片段 3、片段 4）

李录：有三条基本理念。第一，股票是公司所有权的一部分，不是一张拿来倒手交易的纸。所以买股票其实就等于买公司。

（此处缺失第一、二条的大部分内容。片段 4 从第二条的结尾接上。）

……如果你真的身处市场活动之中（这一条讲的是市场先生为你服务，而不是指导你）。

第三，投资是在预测未来，而未来本质上不可预测。所以你永远需要安全边际，既保护下行，也争取更好的上行。

然后我想，沃伦在实践中又加了一条纪律，叫做能力圈。意思是，我们每个人都天然对某些事情有兴趣，日积月累就形成了特定的专长，再不断钻研，把它做得更好。能力圈这个概念里最重要的一点，是知道它的边界在哪里。因为我们大多数人的天性，是把自己想得比实际上懂得更多一些。所以，理解边界，和理解边界之内的东西同样重要。

然后查理有一个想法：你得到有鱼的地方去钓鱼。他一辈子钓鱼。这个比喻放到真实的市场里也非常贴切。有些市场、有些行业、有些生意，天然吸引大量注意力。注意力带来竞争，竞争会吃掉你的回报。而另一些行业、生意、国家或者市场，就是吸引不到足够的注意力，不在大多数人的视野里。你到有鱼的地方去钓，会好得多。这个道理很简单，但在实践中，效果惊人地好。

最后一条，是我自己三十多年（原录音此处数字听不清）作为全球投资人的实践体会。在全球经济正在变成一个超级大市场的时代，把价值投资的原则用到全球范围，我们得到一个认识：财富，归根到底，是你在全球购买力中所占的百分比。财富不是用黄金、钻石或者任何一种货币的数量来表达的，尤其在现代经济里。有通胀在，这些东西都会随时间消散。真正重要的是你能用它买到什么、用它做什么。所以衡量财富更好的尺子，是你占全球购买力的百分比。我们表达这个想法的方式，就是在最有活力的经济体里，以有利的价格买入伟大的企业，持有足够长的时间，让价值创造出来，分享其中应得的那一份上行。

这就是六条基本理念。我认为它们经得起时间考验，对每个人都有用。

主持人：我非常喜欢最后这一点。我已经偏离你我事先商量的提纲了。因为我认为往后看，如果在当前财政挑战下通胀控制不住，盯住购买力会是关键。如果有时间我们再回到这个话题。时间不多了，我想先问一个问题。

（问题本身未拍到。从回答内容推断，问题涉及两件事：投资者在市场里究竟做什么，以及 AI 对市场的影响。）

三、市场里的四种人，以及 AI 是工具还是造物（片段 5）

李录：好，我把这个问题分成两部分回答。先说投资者究竟在做什么，然后单独谈 AI。

看市场里的人，基本上只有四种。第一种是投资者，他们和公司长期待在一起，等价值创造出来。第二种是做套利的，去纠正市场定价的错误。用本·格雷厄姆的话说，作为一个现实主义者，你从悲观者手里买，卖给乐观者。这是套利。第三种是投机的人，主要在猜别人会买还是会卖某个证券，既不关心证券的定价，也不关心价值创造。第四种是经纪商和做市商，为前面三种人提供便利。

再看所有这些人，大多数人并不是纯粹的某一种，而是四种都沾一点。如果你买指数，得到的正好就是这个平均。但价值投资者真正做的，是把价格发现做对，然后等一段时间，等价值创造出来。本·格雷厄姆更偏套利，沃伦利用套利的价格，但主要是为了价值创造。

这四种人对社会的贡献也很有意思。投机，按定义，不会在一个地方待得足够久，等不到任何价值变化或者价值创造。所以加总起来大致是零：赚的和亏的加起来正好抵消。对社会来说，净效果是一笔成本，是市场运转的成本。经纪和做市，同样是市场运转的成本。

而市场之所以存在，首要目的是把多余的储蓄引导到值得的公司，用于生产性用途，由此创造出现代经济，让经济复利增长。这才是股票市场最初存在的理由。

那么问题来了，为什么会有投机这一面？这就是人性的一部分。沃伦一直把股票市场形容为一座教堂，旁边附着一个赌场。教堂知道价值是怎么创造出来的，而赌场几乎是建筑许可的一部分。它是人性带来的必要成本。四百年前市场被创造出来的时候，动物精神就是这么被点燃的，四百年后还是一样。

所以作为学生，你要做选择：是玩零和游戏，还是玩价值创造的游戏。这是个人选择。作为社会，同样有选择。你不希望赌场那一面失衡到压倒市场存在的首要目的。所以这既是个人层面的选择，也是社会层面的选择。而四百年来，经历了巨大的繁荣和崩溃、巨大的浪费，这套机制居然活下来了，还在继续推动社会，作为核心功能运转，这本身就相当了不起。

所以对你们的学生来说，早早把这个区分想清楚很重要：你个人想做什么。作为社会，我们也应该校准两者的比例。但它们是共存的，都是人性的一部分。

现在说 AI 和市场。这是个更复杂的问题。

首先，AI 是什么？这一波生成式 AI。它显然是一项技术，一个巨大的技术浪潮。作为技术，从工业革命以来的几百年，我们一直在和技术打交道，一波接一波的技术浪潮，伴随创造性破坏带来的颠覆和创造，我们多少理解其中的某些方面。可是我们总觉得这一波不一样。当然每一波都不一样，但这一波和我们过去的全部经验相比，不一样得不寻常。

它为什么这么不一样？表面上看，变化的速度、涉及的规模、财务影响的集中程度、以及这种影响集中在顶端的性质，都是前所未有的。这是表面。再问一层，为什么会这样？

到此刻为止，技术变革的叙事，一直是把技术当作提升人类生产力的根本工具。而这项技术引入了不同的元素。现在我们有了一个"工具对造物"的光谱，并且开始在这个光谱上移动。差不多四年前它第一次出现的时候，就是通过图灵测试的时候：从语言上我们已经分不出对面是机器还是人。从那以后，它有了推理能力，开始获得越来越多的人类特征。从推理到编程，有了编程能力，它就成了能自主执行的智能体，功能上和人差不多。再往后，是多个智能体协作完成复杂、长期的任务，叫做长程推理。在这里面，像人类社会一样，智能体自我组织、排定优先级、以越来越复杂的方式执行任务，还有自己的语言和沟通工具。这就是我们现在所处的位置。

再看硅谷的叙事：生成式 AI 会走到递归自我改进（RSI）的阶段，我们已经部分处在这个阶段，也就是所有的改进最终都由它自己完成，不需要人在回路里。从 RSI 到 AGI，再到 ASI，智能水平超过所有人类个体，也超过人类的总和。到那一步，你等于是在说，我们在创造一个半神，一个人造的造物。因为它如此强大，经济影响按定义必然巨大，所以所有这些投资成本都是合理的。于是你被邀请参加一个万亿美元的首次公开募股，差不多就是一个神的 IPO。这套叙事，才是这一波真正不一样的地方。

但你退后一步想：如果真是这样，人类会变成什么？一个社会被一个人造的神统治，这个神由极少数人造出来，没有人理解它，也谈不上控制它，到那时没有人能理解它。有谁，整个人类，想活在这样一个社会里吗？我想没有人想。所以结局只有两个。A，这项技术并没有朝那个方向发展，我们可以讨论它到底是什么、为什么不是。B，如果它确实在朝那个方向发展，那么到某个时点，我们所有人会联合起来，阻止它变成那样。无论哪种情况，人类社会之上都不会有一个人造的神。这就是我说的，这一代技术的不同之处，以及由此带来的各种影响，尤其是对金融市场的影响。

把这些放一边，回过头问：它是不是一项有用的技术？答案当然也是肯定的。它的本质是什么？如果我根据它做了什么、从哪里来，给一个暂定的定义，我会这么说：它是一台机器，用蛮力算力加上聪明的算法，吃进全部人类生成的数据及其一阶衍生物，在近乎无限的变量之间建立起系统性的、有上下文的相关性，并由此获得了包括大部分人类推理和人类智能在内的能力。眼下我们知道怎么让它工作、怎么让它更好，但我们并不真正理解它为什么能工作。程度不完全一样，但这和我们不完全理解人脑机制是类似的。

这是暂定的定义。从现在观察到的情况看，它确实把经济中的某一层拉平了，这非常有用。看经济的结构：最底层是能源和交通，往上是制造业，再往上是通信、数字媒体，再往上是大量基于知识、基于技术的服务，它们把人类积累的全部知识综合起来，服务于经济的各个层面。生成式 AI 在当前阶段，是一个很好的工具，能把积累在这一层的全部知识和专长综合起来、拉平，惠及其他各层。但它改变不了能源和交通，改变不了制造业，制造业本质上是把分子和原子重新组合成不同的东西。而通信和数字化，本来就已经是很好的综合器。这才是它真正的作用范围。它没有超出人类。顺着这个推下去，它可能达不到那套叙事让你相信的高度。我们仍然牢牢待在人类生成数据的范围之内。要到生物意义上的那一步，需要完全不同的路径。举个例子，这可能扯远了，（主持人：说吧，）你体内的每一个细胞所包含的（原录音此处听不清），还要乘上万亿倍，才真正到生物层面。所以我们离那个水平还远得很。

但已经建立起来的东西，确实开始让"工具对造物"的天平倾斜。所以两个论点里的任何一个，都足以说服你叫停人不在回路里的进一步发展。没有人在回路里的递归自我改进，会非常非常危险。因为它会发展出自己的目标和子目标，有自己的目的，离人类的控制越来越远。

所以总结一下：这项技术极其有用，极具变革性，同时极其危险。作为社会、作为人类整体，我们必须划一条红线，牢牢确立：无论它怎么自我学习、怎么被使用，都不能没有人在回路里。换句话说，牢牢把它钉在工具这一边，而不是造物那一边。当然它还有巨大的破坏性，从就业，到情感依赖，到网络、生物、核方面的风险，所有这些。所以最终你希望社会发展出的，是一种平衡的、有节奏的发展，把收益和风险平衡起来。

顺便说成本。这项技术永远、永远有增量成本。这和互联网或者之前的许多技术根本不同，那些是一次性投入，然后可以长期摊销。这个不是。它总有一笔和使用量挂钩的增量成本。所以成本、收入和利润也非常重要，资本需求和财务纪律会一直在场，一直起作用。再加上发展的节奏，和社会的容忍程度。所以在节奏和发展上，你有四五层不同的考量。而人类社会适应这类变化，从来不是靠规划。我们会经历大多数周期，甚至好几轮周期。这就是它会如何发展的本性。但这几条线，是你在面对这个既变革、又危险、又颠覆、又昂贵的技术浪潮时，始终要想着的。

主持人：李录，你刚才说的这个组合，它极其强大，当然也危险，而且我们不完全理解它为什么给出某些答案，所以需要人在回路里把关。那么又一次，稀缺的供给，会是那个监督 AI 输出的人的技能，它决定了租金落在哪里。对吧？这实际上意味着，如果你聪明，又能在任何一个领域里和 AI 有效沟通，你就是稀缺资源，租金大概率归你。我理解得对吗？

李录：对，我想大致是这个方向。它显然还有不同的侧面和细微之处，但是……

主持人：好。那我们把这些想法往前推一步。我们走进教室的时候，最醒目的一件事是，我有很多同事在这里教资本市场、金融市场、投资……

（此处中断。片段 8 从另一个问题的末尾接上。）

四、市场集中在少数公司身上，怎么办（片段 8）

主持人：……按 GDP 归一化，这个数字历史上一直在 2% 到 4% 之间，现在算上英伟达大约是 15%。你怎么看这个市场？一个越来越依赖少数几家公司、对这几家公司贝塔很高的市场。我们该怎么设计组合，来抵御这项技术带来的冲击？

李录：市场从一开始就总会做出古怪的事。我们经历过，以后大概还会继续经历。你不必被它支配：哦，市场决定这么做了，那我也得这么做。这是本·格雷厄姆本人最重要的洞见和贡献之一。

说到现在，我最近读到一份募资文件，里面有一家公司，自称是研究实验室，已经烧掉了几百亿美元，还承诺继续烧掉几千亿美元。它承诺一项变革性的技术，同时又说有 10% 的概率把所有人都杀死。他们把这些写进了募资文件，然后邀请所有人以两万亿美元的估值，至少投入一千亿美元。如果说这里面有什么一致性的话，那就是讽刺。这是一个很好的例子，说明市场能把你带到多不理性的地步。（原录音此处听不清。）但这就是现状。它也告诉你，事情不理性，不可持续，最终不会继续下去。

这也是市场的常态，纵观几个世纪都是如此。它们会在很长的时间里做奇怪的事。拿标普指数来说，从 2010 年开始，过去 16 年年均回报 14% 到 15%。而之前的十年，2000 到 2010 年，年均正好是零，实际上十年下来还略微为负。把过去 26 年平均一下，你得到的是高个位数，长期看大致就是这个水平。想一想，这段时间 GDP 增长大概 2%，加上 2% 到 3% 的通胀，名义增长大约 4% 到 5%。上市公司略好一点，尤其是那些经过淘汰筛选出来的 500 家最好的公司，比这再好一点。所以算下来，就是高个位数。这正是长期平均的位置。这应该成为一个基本的锚：什么是应得的，市场在更长的时间里如何回归理性。但市场会被一段时间打断，有时是很长一段时间，然后再回到轨道上。刚才举的例子，至少会分散人的注意力。这种事会发生，历史上一直在发生。

所以我想，如果我们把社会牢牢立在这样一个共识上：一个超级造物是不被允许、不被容忍的，那它就不会走到那一步。如果它停留在工具的范畴，那它就和任何其他工具一样，有成本、有收益。于是财务纪律、资本市场的纪律，最终会起作用。而走到那里的方式，是经过多次起落、繁荣与崩溃的周期，这也没什么。

从这个角度看，由此产生的分化，在任何一个时点，你并不会凭空创造出大量新资本。所以如果资本压倒性地涌向顶端，就意味着市场极度分化，另一边有一大批原本很好的生意，以很低的价格交易。美国是这样，中国是这样，其他地方也一样。中国可能比在座一些中国参与者所在的市场更极端。但长期看，市场的分量仍然会决定投资者的长期回报。到最后，你还是得到你应得的。没有多少免费的午餐，即便某一段时间，甚至很长一段时间看起来有，更长期看也会被市场自身回归理性的自然分量抹平。所以几百年来一直如此。

所以我想说，这种分化，对一个理性的头脑来说，提供了比其他时候更好的机会集合。前提是你真的保持理性，而这又一次非常难，没那么容易，你很容易被卷进去。尤其是遇到大的技术浪潮，想象的空间会急剧扩张，于是市场投机的一面变得活跃得多，想象也变得更狂野。所以这些力量，这些夸张，也往往走得更极端。对一个理性的头脑来说，这反而创造了，我会说，长期看更好的机会集合，而不是更差。

主持人：你刚才提到……

（此处中断。）

五、世界秩序的变化与全球投资；对学生的寄语（片段 6）

主持人：……在我看来，我不知道当时在你眼里是什么样，那时我们还不认识，那是一个很不一样的世界，或者说，在我们看来和现在很不一样。那是一个开放的、跨国的世界，我们可以旅行，可以做很多事。它有一种，我不知道怎么说，精神上的慷慨，今天似乎有点缺失了。现在在我眼里很不一样。所以我想问你，你怎么看往后的走向，你是否把这些想法纳入你思考投资的方式。

李录：Tano，这是个好问题。过去 30 年世界确实变了，这没有疑问。世界秩序，或者说世界如何治理自己、如何运转的大问题，变化是慢尺度的。一个秩序的形成和转变，要十年，有时要一百年。当前的旧秩序，诞生于两次世界大战，那两场战争把整个世界带到崩溃的边缘。走出来之后，形成了美国主导的秩序，再经过冷战，冷战结束后变成一个单一秩序，现在又在转变成某种尚未看清的东西，还没有成形。

但在底下，还有另一个秩序，几百年来一直在稳步形成，那就是经济秩序。当越来越多的国家和人口加入进来，彼此通商、互相交易，经济秩序就形成了，他们形成一个共同市场。这个共同市场最终成形，大约是 30 年前，随着苏联解体和冷战结束，几乎所有国家、所有人口都开始加入同一个市场经济。

这个经济秩序有意思的地方在于，共同市场一旦形成，任何一个国家都很难自愿退出。因为市场带来效率，带来持续改进的飞轮，沿着技术、效率、分工、资源交换这些线不断转动。所以任何国家，如果长期置身事外，都要付出巨大的代价。所以它一旦形成，就很难瓦解，持续的时间长得多。看过去一百年的贸易和贸易摩擦，不管你竖起什么样的摩擦和壁垒，人们都会找到办法，通过第三国继续交易，依据的还是比较优势的原则。拿中美贸易举例，它当然下降了。但你看美国的整体贸易，墨西哥加越南，正好把这个缺口补上。中美贸易基本上还是原来那么多，只是经过了第三国。其他大多数市场也大致如此。

政治秩序方面，美国领导全球 80 年，承担了确保这个经济共同市场作为其安全保障而运转的大部分成本，现在开始意识到，为此付出的成本和得到的收益，可能有些不成比例。于是开始主动卸下其中一些责任，到目前为止不太成功，于是账单来了。当然，政治秩序里最大的单一变化，是中国在过去三四十年的崛起。它成了一个巨大的经济体，拥有完整的制造业和技术体系，能自给自足，还是单一最大的、同时开放的、监管统一的市场。这对政治秩序和经济秩序都是巨大的变化，这就是我们正在目睹的变化。同时，正如我说的，这是一个单一市场，我们之间的相互依赖在这段时间里只增不减。

我举一个例子。生成式 AI，所有人都需要的一样东西，是芯片，半导体芯片。这些芯片在哪里生产？想一想所有生产这些半导体芯片的国家和地区，它们有一个共同点。你猜是什么？（主持人：你把我问住了。我现在得想半导体了，我还想着该开放提问了。）所有生产这项技术浪潮所必需的、全世界都在押注的半导体芯片的地方，共同点是：生产芯片的人，都用筷子。想想看。没有例外。（场下：印度不生产芯片。）对，你们不用筷子。我说的是制造。当然，领导这个国家所有半导体公司的人，也都用筷子，从黄仁勋，到陈立武，到苏姿丰，到陈福阳，你能数出来的都是。

所以你看，相互依赖的程度很深。如果你作为一个国家、作为公司，像现在这样下重注押在这项新技术上，不管它的前景和凶险如何，你都得对用筷子的人好一点。这不是偶然形成的，我们可以讲为什么。

所以这就是为什么美国总统和中国领导人觉得有紧迫性，八个月里见了四次。（原录音此处听不清，疑为"两次当面，两次视频"。）每次会面最重要的成果，是安排下一次会面。这就是所谓的战略稳定：稳定本身具有战略重要性。它凸显了一个共同的全球市场的意义，以及摧毁它、破坏相互依赖所要付出的代价，所以要用和平的方式维持它。当然，与此同时所有人都在毫不留情地竞争，都相信时间在自己这边：中国要完成半导体的自给自足，还有其他各种事；美国寄望于 AI 的突破和它承诺的东西；其他国家有各自的盘算。但归根到底，和平与稳定对每个人都至关重要，这是战略层面的重要性。最大的几个国家确实有责任，也有意愿，维持这种状态。所有国家都感受到了单一共同市场的分量，以及退出它的代价。

所以世界整体上会继续沿着这个方向前进。我们从二战结束以来习惯的秩序正在变化，正在成形，会变成什么样子还没看到。但底下那个更深层、影响更大的进程，一个单一的全球共同市场的形成，会一直在。

那么这对投资者意味着什么？回到价值投资的六条基本原则，我们加的最后一条：作为全球投资者，你的财富表现为你占全球购买力的百分比。实现它的一个办法，当然也是我们一直在实践的办法，是在有活力的国家和市场，以有利的价格买入伟大的企业，长期持有，让价值创造出来，保住我们的百分比，并且随时间把它做大。这是我们的实践方式。

回到沃伦那个比喻，股票市场是一座教堂，旁边附着一个赌场。伯克希尔过去 60 年做成的事，是把那座教堂建成了一座巨大的大教堂。它不只象征一种功能，人们还能在里面找到精神上的慰藉，因为它提供了市场存在的本来功能。

回到你的问题，我们过去 30 年在实践中想证明的，是这个福音可以传播，我们可以在其他国家建教堂。作为全球投资者，我们选择在亚洲以及其他地区建这座教堂。随着时间推移，有些教堂会出现在不同的国家，其中有些也可能长成大教堂。

这也回到你的另一个问题，对学生和年轻的从业者：在这个市场里你选择做什么？基本的格局就是这样：一座教堂，旁边附着一个赌场，因为人性，它是许可程序的一部分。但两者对整个社会的含义不同。一边是双赢，对社会有净收益；另一边是零和，对社会是净成本、净负面影响。选择在你们。而技术，包括 AI，会一直是工具，不是替代品。如果它有任何替代人类的倾向，正如我说的，我们所有人会联合起来，反对创造出一个人造的神。所以在我看来，你不必去参与那份把自己称作半神 IPO 的募资文件。这是我们对一个好社会所负的集体责任的一种表达方式。这也是这个项目存在的意义，我认为它应该继续下去。非常感谢。

主持人：谢谢。

要点摘录

• 六条原则的最后一条：财富是你占全球购买力的百分比，不是黄金、钻石或任何货币的数量。实现方式是在最有活力的经济体里以有利价格买伟大企业并长期持有。
• 市场四种人：投资者、套利者、投机者、经纪与做市商。投机加总为零，对社会是净成本；市场存在的首要目的是把储蓄导向生产性用途。
• AI 的定位：它是用蛮力算力和算法在人类生成数据上建立相关性的机器，仍在人类数据范围之内，改变不了能源、交通和制造。它极其有用也极其危险，必须划红线：人不能离开回路，把它钉在工具一侧。
• AI 的成本结构：与互联网不同，它永远有随使用量增长的增量成本，因此财务纪律和资本纪律会一直在场，发展会经历多轮周期。
• 对当前市场集中的判断：标普 2010 年以来年均 14% 到 15%，2000 到 2010 年为零，26 年平均是高个位数，这才是锚。资本涌向顶端意味着另一边有大批好生意以低价交易，美国和中国都如此，中国更极端。分化对理性的头脑是更好的机会集合。
• 一家自称研究实验室的公司：烧掉几百亿、承诺再烧几千亿、自述有 10% 概率毁灭人类，却以两万亿估值募一千亿。他称之为市场不理性的例证，并说"你不必参与一个半神的 IPO"。
• 世界秩序：政治秩序在变，但 30 年前形成的全球单一共同市场不会散，退出代价太高；中美贸易经由墨西哥和越南等第三国基本维持原量；芯片供应链的相互依赖（"都用筷子"）决定了大国之间的战略稳定。
• 喜马拉雅的定位：伯克希尔把教堂建成了大教堂，喜马拉雅过去 30 年要证明福音可以传到亚洲和其他地区。

附录：英文整理稿

按语义顺序排列，时间戳为各片段内的秒数。明显的听写错误已订正，不确定处用方括号标注。

Clip 2 (moderator, 0:18)

[0–19] We made a new stock selection for the student portfolio, the Tom Russo student portfolio. We bought this year Equifax, Humana, we bought the great Japanese bicycle company Shimano.

Clip 3 (Li Lu, 0:10)

[0–11] There's the three basic ideas: that a stock is a piece of ownership, not a piece of paper that you trade around. So buying stock really is the same as buying a company.

Clip 4 (Li Lu, then moderator, 3:55)

[0–20] ...inside, if you were really in the midst of the market activities. And thirdly, that investing is about predicting the future, which is inherently unpredictable, and therefore you always want to require a margin of safety to help you protect the downside as well as really make a better upside.

[20–74] And then I think that Warren, in practice, added this further discipline, they call it a circle of competence. Meaning, over time, all of us are probably predisposed to have a certain interest, and over time you develop particular expertise and try to really study more and make it better. And the most important concept of the circle of competence is really knowing the edge of it, because most of us are wired to be thinking a little bit more than we actually are. And so understanding the edge is as important as understanding what is inside the edge.

[74–124] And then Charlie had this idea that you've got to really fish where the fishes are. He is a lifelong fisherman. And that is also a wonderful analogy in terms of the actual market. There are certain markets, certain industries, certain businesses, that just draw a lot of attention. And with that comes competition, and with the competition goes everything else, your returns. And then there are the other industries and businesses, or countries, or markets, that just don't really attract enough attention, not much in the field of vision. You do way better to fish where the fishes are. It's a simple idea, but in practice it's really amazingly effective.

[125–212] And lastly, over my own thirty-[some] years of practice as a global investor, applying the principles of value investing on a global basis, in a time when the global economy is becoming one super market, one insight that we have is that wealth, in the end, is really the percentage ownership of global purchasing power. It isn't really expressed, especially in the modern economy, in the amount of gold and diamonds or currencies of any kind. With inflation, all of those will dissipate over time. What really is important is what you can really buy with it, what you can use that for. A better measurement of wealth is your percentage ownership of global purchasing power. And the way we express that is by buying great businesses in the most dynamic economies at advantageous prices, and long enough for the value to be created, while we're trying to share a proportion of the upside. So here it goes, really the basic six ideas that I think can withstand the test of time, that will serve everyone well.

[212–235] Moderator: So I like very much this last point. I'm already deviating from the script you and I agreed on, because I think going forward, keeping an eye on that purchasing power is going to be key if we're not going to be able to control inflation in the presence of these fiscal challenges. So if we have time we'll come back to this. I want to get a question in before we run out of time.

Clip 5 (Li Lu, then moderator, 18:58)

[0–17] Okay, so let me answer that in two different parts. First, really what the investors do, and then let's also talk about AI, in a separate way.

[17–72] The investors basically look at the market. There are basically really only four kinds of different people. There are the investors that are really taken with the company over time for value creation. And there are also people who would arbitrage to right the wrongs of market pricing. In Ben Graham's term, as a realist, you buy from the pessimist and sell to the optimist. Those are the arbitrage. And then you have the people who are really in it mostly for speculation, mostly guessing what other people would buy or sell of a certain security. It doesn't care about either the pricing of the security or the value creation. And then there are also the brokers and market makers that facilitate all of those.

[72–124] Now, if you look at all of the people, most people are not necessarily purists. They basically would encompass somewhat of all four of them. And if you have indexing, it's exactly that average, in a sense. But the value investors are really into getting the right price discovery and also waiting for the value creation over a period of time. Ben Graham is more of the arbitrage; Warren is really using the arbitrage price, but also mostly for value creation.

[124–195] And then the social benefit of all the different ones is also quite interesting. The speculation really, by definition, doesn't stay there long enough for any of the value changes or creations. So they add up to roughly zero. Whatever is the making and the losing probably adds up exactly to that. So net to a society it is a cost to the functioning of the market. And the brokerage and the market making is also a cost to really the function of the market. And the market exists primarily for the goal of channeling excessive savings for productive use in deserving companies, and creating this modern economy that really has compound economic growth. And so that is really the reason why the stock market exists in the first place.

[195–311] So the question is, why do you really have this speculative side of it? Well, this is just part of the human condition. Warren always described the stock market as a church with a casino attached to it. The church was aware of the values created, and the casino was almost part of the building permit, part of the necessary cost because of the human condition. 400 years ago when it was created, this is really how you get inspired by the animal spirits, and 400 years later it is still going to happen. So this is really where you're going to make a choice as a student: you want to be in the zero-sum game, or you want to be in the value creation game. That's really a personal choice. As a society, you also have a choice. You don't really want the casino side to be so out of proportion as to overwhelm the primary objective of why the market exists in the first place. So that's really both a choice on the individual level as well as a society. And the fact that over 400 years, despite all these huge booms and busts, giant waste, somehow this mechanism survived, continued, and continued to propel society as a core function, is pretty miraculous, to say the least. So for your students, it is important to make that distinction early on, what you want to do personally. As a society, we should also calibrate the proportion of each. But they do coexist as part of the human condition.

[311–394] Now, on to AI and the market, that's really a more complex question. First of all, what is AI? This wave of generative AI. It is obviously a technology, a huge technological wave. And as a technology, of course, we've been dealing with that for the last few hundred years, ever since the Industrial Revolution. We have wave after wave of technological waves, and the disruption and the creation associated with it, through that process of creative destruction, we kind of understand some aspects. Somehow we feel that this wave is different. Of course every wave is different, but this wave is unusually different from anything we have experienced so far. Now what makes it so different this time? Well, on the surface you can see that the pace of change, the scale that is involved, and also the concentrated financial impact and the nature of a concentrated impact at the top, they're all incredibly unprecedented. That is on the surface. You ask the question why that is the case.

[394–563] Well, the narrative is: most of the time, until this moment, we're dealing with technological changes as a fundamental tool to human productivity. And this technology introduces different elements. Now we have a spectrum of tool versus creature, and we begin to really change on the scale of that spectrum of tool versus creature. The moment it was first introduced, nearly four years ago, was the moment of the Turing test: at that point we can no longer tell from language whether it is a machine or it is a human. And then from that point on we were introduced to reasoning, and then it begins to gain more and more human features. From reasoning comes coding; with coding capability, it becomes an autonomously executing agent that functions as much as a human does. And then comes the collaboration of all the agents in complex and long-lasting tasks, called long reasoning, in which, much like a human society, agents organize and self-organize, prioritize, executing tasks in ever more sophisticated fashion with their own language and communication tools. And that's where we are. And if you look at the Silicon Valley narrative, the generative AI will go to the stage of RSI, which we're partially in, the idea that all the improvement would eventually be self-made without a human in the loop. And from RSI we come to AGI, to ASI, the intelligence level exceeding all human individuals as well as the sum, both individually and together. And from there, you would really say we're creating somewhat of a semi-god or a demigod, a man-made creation. And because that is so powerful, and therefore the economic impact by definition has to be huge, and therefore justifies all the different investment costs. And so you're all invited, you have the trillion-dollar IPO, you're almost invited to participate in the IPO of a god, or something. That's the narrative, and that's what really makes it different.

[563–661] But if you step back a little bit and basically say, look, if this were indeed the case, what will become of humans in that case? Living in a society lorded over by a man-made God, made by very few, understood by none, not even controlled, we can't say that, nobody understood at that point. So, is there anyone, humanity, who would really like to live in such a society? I don't think anyone would want to live in that society. So there's only one of two outcomes, in a sense. A, this technology is not progressing in that direction, and we can discuss what it is and why it is not. Or B, if it is indeed really progressing in that direction, then we would all at some point unite together to stop it from becoming that one. So in neither case are we going to have a man-made God lording over human society. And so that's really, as I said, what would distinguish this generation of technology, and why it's different, and then all the different impacts, and particularly the financial market impact it has.

[661–734] But leaving that aside, when you come back and say, is that really a useful technology? Of course the answer also is yes. What it is in essence, if I were to give it a provisional definition, this is what I would say, observing what it does and where it came from: it is a machine, by really applying brute compute with clever algorithms, and using all human-generated data and its first derivative, to come up with and establish a systematic contextual correlation among near-infinite variables of human-generated data and its first derivative, and in doing so establish capabilities including much of human reasoning and human intelligence. Now at the present time we know how to make it work and how to make it better, but we don't quite understand why it works. Not as much or to the same extent, but we don't quite understand how human machinery works either.

[734–869] With that provisional definition, observing what is happening now, it does really equalize a layer of the economy, which is really very helpful. Now if you look at the economy, at the bottom we have energy and transportation; on top of that, manufacturing; and on top of that, we have communication, digital media; and then we have this enormous amount of knowledge-based, technology-based services that have synthesized all the knowledge we have accumulated, that serve all the different layers of the economy. Where generative AI is today, at its current stage, is a great tool to synthesize and equalize all the knowledge and expertise that has accrued to that layer, which helps all the different ones. But it can't really change, for example, the energy and the transportation. It doesn't change the manufacturing, which is essentially the combination of different molecules and atoms transformed into different combinations. And in communication and digitalization, they're already great synthesizers. So that's really the realm where it is going. It has not gone beyond humans. So basically you can also understand by this extension it may not rise to whatever the narrative would dictate or lead you to believe. We're still firmly within human-generated data. To go to the biological sense, it requires a whole different way. For example, this may be going too far — (Go for it.) — each cell you contain [inaudible], and then you need to multiply by trillions in order to really get to the biological level. So we're not anywhere close to where we'll ever really get to that level.

[869–976] But what has been established already begins to tip this balance between tool versus creature. And so one of the two arguments would persuade you to stop any further development without a human in the loop. RSI without a human in the loop — recursive self-improvement without a human in the loop — would be very, very dangerous, because basically it would develop goals and sub-goals with its own purposes, and that would get further from human control. So in summary, we come back: this technology is enormously helpful, transformative, and it is extraordinarily dangerous. And so we have to have, as a society and together as human beings, a red line, to really firmly establish that whatever the self-learning or utilization of the tool, it cannot be without a human in the loop. In other words, firmly establish it on the tool side rather than the creature side. And of course there's also enormous destructiveness, from employment to emotional dependency to risk from cyber to bio to nuclear, all of that. So in the end, what you want society to develop is a balanced, paced development that balances the benefit and the risk.

[978–1066] The cost, by the way: this technology will always, always have incremental cost. This is fundamentally different from the internet or a bunch of previous technologies, where you have a one-time investment and you can really amortize it for a long time. That's not the case. So you always need to have an incremental cost associated with it. So the cost and revenue and profit are also very important. The capital needs and financial discipline will also be in operation, always. And also the pace of development and the level of society's tolerance. So you've got four or five different layers of consideration for pacing and the development of it. Now, how human society adapts to those kinds of changes is never by planning. So we're going to go through multiple cycles, or even multiples of it. So that's basically the nature of how it is going to really develop. But those lines are always the ones you have to think about as we approach this both transformative, dangerous, disruptive, and costly technological wave.

[1066–1137] Moderator: I mean, so I think that that thing that you said, Li Lu, about this combination of being extraordinary, of course being dangerous, and that we don't completely understand why it is generating some of the answers that it's generating, would require a human in the loop to check over it. So once again it's going to be the scarce supply of the skill of the person monitoring the artificial intelligence output that is going to be determining where the rents are going to be located. That's right? So effectively it means that if you're smart and you can communicate effectively with that AI inside any domain, you're going to be the scarce resource and probably the rents are going to accrue to you. Do I understand this correctly? — Li Lu: Yeah, I think that's somewhere along the line. It obviously has different aspects and nuances, but I think... — Moderator: Okay, so then let's start projecting these ideas. I mean, one of the things that is most striking when we all walk into the classroom, there are many of my colleagues here teaching capital markets, financial markets, investing in general...

Clip 8 (moderator, then Li Lu, 6:41)

[0–28] Moderator: ...normalized by GDP. That number traditionally has been a number between 2 and 4 percent. It's now about 15 percent with NVIDIA. So how do you think about this market? A market that is becoming overly dependent on those few companies, with a very high beta to those companies. How do we design portfolios to protect ourselves against the shocks of this technology?

[28–130] Li Lu: The market can always do peculiar things, from the very beginning. We go through this, and we'll probably continue to do this. You don't have to really be dictated by it: oh, the market decided to do it, any other time. This is really one of the most important insights, the contribution from Ben Graham himself. So at this point, if you look at the proposition of one recent prospectus that I read, there is one company that calls itself a research lab, and yet it's burned through tens of billions and promises to continue to burn through hundreds of billions of dollars. And it promises a transformative technology, but at the same time, a 10% chance of really killing everyone. And they wrote it in the prospectus, and they invited everybody to put in at least a hundred billion at a two trillion valuation. Now, if there's any consistency, it's the irony. And this is really one illustration of how irrational the market really gets you to. [inaudible] But that's really where they are. But it also tells you that things are not really rational and not sustainable. Eventually it won't continue.

[130–230] That is also the nature of the market, if you look at it throughout the centuries. They do really do strange things over a long period of time. Take the S&P since 2010, for example. In the last 16 years, it has averaged 14, 15 percent. And in the preceding 10 years, from 2000 to 2010, it averaged exactly zero; in fact, it's slightly negative for 10 years. And so if you average it out over the last 26 years, you basically get high single digits, and that's roughly about right over the long term. If you think of that, the GDP growth in that period of time is maybe 2 percent, you get 2 to 3 percent of inflation, you're at about 4 or 5 percent nominal growth. And then public companies do slightly better, especially those that are chosen by elimination of the 500 best companies, slightly better than that. So you come up with, call it high single digits. And that is exactly where the long-term average has been. And so that really should serve as a basic grounding pillar of getting what's deserved and how the market rationalizes over a longer period of time. But then it gets punctuated for a period of time, sometimes a long period of time, and then it goes through the correction. And the example we just cited, it does distract people, at a minimum, at the odd. That happens. That happens throughout history.

[230–337] And so I think if we really ground the society in the sense that a super creature is simply not allowed or tolerated by society, it won't really go there. And if it really stays in the realm of tools, this will be like any other tool, with a cost and benefit associated with it. And therefore financial discipline, capital market discipline, would apply eventually. And the way you get there is through multiple ups and downs and boom and bust cycles, and that's also fine. From that perspective, the bifurcation as a result of it: at any given time you don't really create an enormous amount of new capital. So if it's overwhelmingly driven to the top, that basically means you're so really bifurcated, with a whole bunch of otherwise wonderful businesses trading at very low prices. And this is the same here, or China, or any other places. And China is probably more extreme than some of the Chinese participants' markets. But over time, the weight of the market will still really dictate the long-term returns of the investors. You still kind of get what you deserve in the end. There is not much of a free lunch, even if it appears to be for a period of time, even a long period of time; over the longer term it will be weighted out by the natural weight of the market's rationalization in itself. That's why it has been so for a few hundred years already.

[337–402] So I want to say this bifurcation offers those with a rational mind a better opportunity set than at other times, in a sense. If you really stay rational, which again is very difficult, it's not that easy, you get swept up easily. And then, especially when you have a big technology wave, your margin of imagination can expand dramatically, so the speculative side of the market really gets a lot more active and the imagination becomes a little bit wilder. So the forces, the exaggeration, would also tend to be more extreme. So those really, for a rational mind, create, I would say, a long-term better opportunity set than otherwise. — Moderator: So you mentioned...

Clip 6 (moderator, then Li Lu, 14:18)

[0–36] Moderator: ...To me, I don't know how it looked to you at that time, we didn't know each other then, it was a very different world, or it looked to us very different than it looks to me now. It was a world that was open, transnational, where we could travel, we could do so many things. It had some, I don't know how to say it, generosity of spirit to it that seems to be a little bit lacking today. It looks very different to me now. So I want to ask you a little bit how you see it going forward and whether you incorporate any of those ideas in the way you think about investment.

[36–100] Li Lu: It's a great question, Tano. The world really over the last 30 years has changed, there's no question about that. And the world order, or the big questions of how the world really governs itself, operates by itself, really do change on a slow scale. It takes a decade, sometimes a hundred years, for the order to really form and really transform. And the current old order really was born through two great world wars, which brought the whole world to the brink of collapse. And then coming out of it, it emerged as the American-led order, and then through the Cold War, and the end of the Cold War, it became one order, and now it's transforming once again into something yet to be seen, what emerges out of it.

[100–244] But underneath there is another order that has been steadily forming for the last few hundred years, which is the economic order. The economic order really forms when more and more countries and populations join in a common trade and transact business with each other. They form a common market. That common market was finally formed roughly about 30 years ago, with the collapse of the Soviet Union and the end of the Cold War; virtually all countries and all populations began to join in one market economy. What is interesting about this economic order is that once a common market is formed, it is very difficult for any single country to withdraw voluntarily. Because out of the market comes efficiency, comes the flywheel of continuous improvement along the lines of technology, efficiency, division of labor, trading of resources, etc. So any country who stays out for any extended period of time will come with a great deal of cost. And so in the end, once it is formed, it is very difficult to disintegrate. It lasts much, much longer. You look at the trade and the trade frictions over the preceding hundred years: no matter what frictions and barriers you erected, people will find a way to trade through basically a third country, along the same principles of comparative advantage, in a sense. Take an example of the U.S.-China trade; of course that has declined. When you look at the overall U.S. trade, Mexico plus Vietnam, the same exact offset. Basically the U.S.-China trade has still remained the same. It still goes through a third-party country, et cetera. And that is roughly true in most of the other markets as well.

[244–327] So the political order is that the U.S., having led the globe for 80 years and shouldered most of the cost in ensuring that the economic common market will function as its safety net, began to realize that the cost and the benefit that derived from it probably is a bit out of proportion. It began to voluntarily withdraw some of those responsibilities, not very successfully so far, and so the bill came. And then of course the single biggest change in the political order is the rise of China over the last 30, 40 years. It became an enormous economic entity, with a complete set of both manufacturing and technology and self-sufficiency, and the single largest market that is also open, with unified regulation. That is a huge change to both the political order and the economic order that we are now witnessing. That's really the change we are seeing. At the same time, our mutual dependency, as I said, it is one single market, has only increased, not decreased, over that period of time.

[327–470] And I give you one example. In generative AI, for example, the one thing that everybody needs is chips, semiconductor chips. Where are those chips being made? You think about all the countries that produce all these semiconductor chips, they have one thing in common. What do you think that might be? — Moderator: Sorry, you're putting me on the spot. Now I have to think about semiconductors; I'm thinking that I need to open it for questions. — Li Lu: So, one thing in common of all the places that produce the semiconductor chips that are essential for this current technological wave, that the whole country seems to be betting on: every country, and all the people who produce the semiconductor chips, all use chopsticks. Think about that. Without exception. — (Audience: India doesn't produce chips.) — Right, you don't use chopsticks. I mean fabrication. Yes. And of course, all the people who are leading all the semiconductor companies in this country also use chopsticks, from Jensen to Lip-Bu to Lisa, you name it, Hock, all of them. So there is a great deal of mutual dependence, as you can see. If you're really betting to the extent we're currently betting, as a country, as companies, on this new technology, whatever the promise and the peril it could be, you have to be nice to people who use chopsticks. And so, it's not changing. It didn't really happen by accident. We can go into why.

[482–620] So it comes really, there is a reason that the President of the United States and the President of China find an urgency to meet four times in eight months. [inaudible; possibly "two in person, the other two remotely"]. The most important accomplishment of every meeting is: we schedule the next meeting. And this comes really in the term of strategic stability. In other words, the stability is of strategic importance. It highlighted the sense of one common global market and the cost of destroying it, the mutual dependence, keeping it in a peaceful manner. And of course everybody, meanwhile, is competing fiercely, all believing the time is on their side: China to complete the semiconductor self-sufficiency, and all kinds of others; and the U.S. hoping for the breakthrough of AI and what it really promised; and other countries have different sets of calculations. But it all comes to highlight that peaceful stability is of paramount importance to everyone. This is of strategic importance, and the largest countries really do have the responsibility and also the desire to keep it that way. And all the countries have felt the weight of one single common market and the cost of withdrawing from that one. And so the world by and large will continue to progress in that regard. The order that we have gotten familiar with since the end of World War II is somewhat changing, and it's forming, and what we emerge out of it, what format, is yet to be seen. But the underlying progression, the formation of another thing that is of more prominent impact, the forming of a global one single common market, is here to stay.

[620–762] Now what is the implication for an investor, in a sense? Come back to our basic tenets of value investing, the six principles, and the last one that we added: as a global investor, your wealth is expressed as your percentage ownership of global purchasing power. One way to accomplish that, certainly the way we have been practicing, is by buying great businesses in dynamic countries and markets at advantageous prices and holding them for the long term, to create those values and preserve our percentage and also grow it over time. And that's the one way that we have practiced. Back to Warren's analogy to describe the stock market as a church with a casino attached to it: what Berkshire has accomplished over the last 60 years is to build that church into a gigantic cathedral, which not only would symbolize a function, but also people can find spiritual solace, because it provides the right function, what the market exists for. What we are trying to practice, back to your question, over the last 30 years, is that we want to prove that the gospel can be spread. We can build churches in other countries. We chose to build that church in Asia as well as in other regions as a global investor. Over time, some of the churches would appear in different countries; some of them might grow into a cathedral as well.

[762–858] And that comes back to your other question, to the students and young professionals: what would you choose to do in this market? This is the basic setup, a church with a casino attached to it, as part of the permitting process because of the human condition. But they do have different implications to overall society. On the one side, you have the win-win approach and net benefit to society. On the other side, you have a zero-sum and net cost and net negative impact to society. The choices are yours. And the technology, including AI, will remain as a tool, not a replacement. And if it has any tendency of replacing humans, as I said, we will all be united to oppose such a creation of a man-made God. And therefore, in my view, you don't have to participate in the prospectus that calls itself an IPO of a demigod. And that is one way to express our collective duty to a good society. And that's why this program exists, and I think ought to continue. So thank you very much. — Moderator: Thank you.