1、两种认知模式的底层差异

【1. Charlie Munger – The Architect of Berkshire Hathaway】

For one thing, I was only interested in whether things worked. Charlie wanted to know how things worked. I would turn on a switch and if the tv or light bulb went on, I could not care less about what had caused that miracle. Charlie, however, would want to understand every aspect of how a generator worked, how electricity traveled to his home, the merits of AC vs DC, whatever. You could say that Charlie understood electricity better than Thomas Edison ever did.

Like his hero, Ben Franklin, Charlie wanted to understand everything, and he pretty well succeeded. 

【批注】“对事物是否有效感兴趣”源自于想解决现实的问题，并且相信“正确答案是存在的”，最后指向的是专注；“理解一切”的出发点和目的地跟前者是不一样的，前者相信知识是收敛的，后者不相信知识是收敛的，即使被告之、被引导、被证明、甚至自己也同意是应该收敛的。

有些“鸟”来到世上是来解决问题的，乔布斯在一次采访中说苹果的成功是专注于“真正重要的事情”，另一些有不同的目的。那些专注于解决问题的会很快发现自己擅长的点，比如，乔布斯发现三星和其他手机的设计师不知道自己是有手指头的，巴菲特在Las Vegas看到很多人喜欢把辛苦赚来的”好钱“扔进“坏钱”堆里。

在发现自己的擅长点以后，他们又会发现自己的时间和精力是有限的，聚焦和专注才能实现来到这个世界的最初的目的。

2、致命的错误

【4. First quarter results and $182B in cash】

And sometimes sometimes, we’ve done things that were big mistakes. But never we never get close to fatal mistakes. And every now and then we do something that really works.

【批注】致命的错误包括：（1）系统性的、直接归零的风险；（2）到期会归零的期权；（3）过高的杠杆，也包括人，会在错误方向加倍努力的人。

3、相比于银行，伯克希尔更有能力吸收新的资本

【4. First quarter results and $182B in cash】

And we have accumulated 571 billion, and I couldn’t help but look at who’s second. And JPMorgan had 327 billion at year end, and they’re up to 338, I believe, at the end of the quarter. But they pay significant dividends. They repurchase shares. They’ve got a business that earns better returns on equity. But they don’t plow it and they shouldn’t they don’t plow it back exactly like we have.

【批注】银行对于新增资本的需求有一些局限性，BRK可以把每一笔钱留下来等待机会，银行还做不到，银行在一个更窄的空间扩张业务、有更多的限制，但是银行也是能够持续扩张的，能够以较高的ROE（15%以上）投入新的资本，JPM就是例子。

4、苹果的三重减值风险

【5. Selling some Apple shares now with higher tax rates expected later】

But investments, you don’t really have to do that. You really have to have your mind set properly. So, we will end up — unless something dramatically happens that really changes capital allocation and strategy — with Apple as our largest investment. But I don’t mind at all, under current conditions, building the cash position. When I look at the alternatives available in the equity markets, and I look at the composition of what’s going on in the world, we find it quite attractive.

And I would say with the present fiscal policies, I think that something has to give. And I think that higher taxes are quite likely. And if the government wants to take a greater share of your income, or mine, or Berkshire’s, they can do it. And they may decide that someday they don’t want the fiscal deficit to be this large because that has some important consequences, and they may not want to decrease spending a lot. And they may decide they’ll take a larger percentage of what we earn. 

【批注】

2、目前的财政赤字难以维持，苹果已经缺少增长的潜力，但是PE仍然维持在30倍以上，未来10年存在3个减值的风险：

（1）资本支出增加、现金流减少；

（2）PE倍数下降；

（3）税收增加。

假设估值的起点是1000亿，PE下降30%就失去300亿，税收从20%提到30%又失去50亿，苹果还可能大幅增加资本的支出，未来10年在剔除这些因素以后还能不能有所增长？

巴菲特认为还是有可能性的，因为这样的好企业不多，目前减仓的主要原因是第1点，持有现金可能有更好的收益。

6、BHE比BNSF的风险更大

【7. Our utilities won't throw good money after bad if states impose excessive climate change costs】

And so whether you earn X or X plus a half a percent or X minus a half a percent, that differs by state. And some states are more attractive than others. But whether you earn X or go broke is not an equation that works. And we won’t put our shareholders’ money — they didn’t give it to us to lose it all. 

【批注】同样是受监管的资本密集型业务，BHE相比于BNSF的风险要大，这个业务会因为火灾赔光所有的资本。巨灾险是BRK非常擅长的业务，无上限的野火索赔是很难接受的逻辑。

7、AI既有好的想法也有坏的想法，不知道哪个先来

【8. Artificial intelligence (AI) could make fraud "the growth industry of all time"】

I mean, scamming has always been part of the American scene. But this would make me — if I was interested in investing in scamming — it’s going to be the growth industry of all time. And it’s enabled in a way — you know, obviously AI has potential for good things, too — but I don’t know how, based on the one I saw recently, I practically would send money to myself over in some crazy country.

it has enormous potential for good and enormous potential for harm, and I just don't know how that plays out.

【批注】一种思考问题的框架，巴菲特用这个方式评价过华尔街的投资经理，既有好的想法也有坏的想法，不知道哪个先来。

8、优势可以掩盖问题，GEICO就是这样

【9. GEICO is catching up with competitors on using data technology to set rates】

But our fundamental advantage at GEICO, of course, is that we have lower costs than virtually anybody. And that cost advantage has been dramatic. We’ve driven our underwriting expense ratio below 10 percent, and there’s — well, there’s just very, very, very few companies that can compete with that. So it isn’t — it’s not in the least a survival question, and it isn’t even exactly a profitability thing. But, you know, we would rather have X percent of the market than a half of X percent.

But our low costs have masked the fact that, for a while, that we could do without progressing as much as we should’ve in the matching of rate to risk.

【批注】巴菲特认为GEICO的成本优势还非常明显，我自己有一些不同的观点。

1、损失率（Loss Ratio，费率与风险匹配的程度），因为技术落后在很长时间里输给Progressive，会不会成为结构性的劣势，可能会也可能不会，前面认为不是关键变量不够重视，现在已经承认错误；

2、直销是不是结构性的优势？有一种可能性是人工智能帮助Progressive去掉中间的保险代理人，Progressive和其他的汽车保险公司都进入以直销为主的模式；

3、GEICO的浮存金可以投资股票，其他的保险在很长时间还是做不到这一点，这是更有确定性的结构性优势。

巴菲特最后指出的点很重要：优势可以掩盖事实，能在大量信息中提炼出关键信息、找出问题边界的人太少，他想到了浮存金的投资收益要跟GEICO的经营分离，但还是在很长时间里让直销掩盖了损失率的劣势。

9、为什么会有不信任的人？

【10. Surround yourself with people you trust, like Charlie Munger】

But if I didn't think I could do it myself, I wouldn't have done it. To some extent I talk to myself on investments.

If you don't live a life where you surround yourself, and limit yourself, to people you trust, it won't be much fun. I mean, I literally have been in the position ever since I was in my twenties, of being able to have people I trusted around me.

【批注】想一想为什么会有不信任的人？为什么会发生？为什么会妥协？是不是因为配不上？

10、保险能承担的是离散的风险，聚合违背了本质

【15. Cyber insurance is hot but too risky for now】

I have told them, no matter how much you charge, you should tell yourself that each time you write a cyber insurance policy, you’re losing money. We can argue about how much money you’re losing, but the mindset should be, you’re not making money on it, you’re losing money. And then, we should go from there.

And you’ve got to have somebody in charge of things that understands that you may get an aggregation of risks that you never dreamt of and maybe worse than some earthquake happening some place just because you have a whole bunch of policies with a million-dollar limit.

【批注】保险能承担的是离散的风险，聚合违背了保险的本质。

11、人生建议：找到目标，保持不变形

【18. Life advice: set goals and take the paths that achieve them】

I mean, find the job that you would like to have if you didn’t need a job. And sometimes you can find that very early. And sometimes you go through various experiences. But don’t forget what you’re actually trying to do.

【批注】"但别忘了，你真正想做的到底是什么"，能在寻找的路途上保持不变形。

12、Greg Abel接班：脑子差一点，手脚勤一点

【19. Greg Abel is now working directly with Berkshire's managers】

Because Greg, in a year, accomplishes I mean, he sees more of them, understands more about their problems, can give them suggestions. He's got incredible amounts of energy. 

Greg will do something about it. And Charlie and I wouldn't have. Not because we didn't know it should be done, but because we were doing so well ourselves. (Laughs) It just wasn't something we'd make the effort for. We didn't want to change our lives that way. 

【批注】巴菲特有些话是非常坦诚的，不去费心Greg Abel的那些事是因为"自己做得太好了"，同样一份时间，做自己的事可以创造更多的价值，Greg Abel显然不具备这个能力，没有能力一大块、一大块捡金子，一粒粒的捡也是捡，脑子差一点、手脚勤一点。

1、找到30岁的巴菲特或者乔布斯，这样的假设一开始就有问题，"自己做得很好的"为什么要去管别人的业务？一块非常大的业务，存在结构性的矛盾；

2、AJIT JAIN在性格特质上更好一些，Greg Abel年龄上更有优势，巴菲特选了后者，说明两个人在最重要的性格特质上没有足够显著的差别。

16、政治上的倾向性：起点是照顾穷人

【24. Settlement may change commission structure, but real estate agents still have important role】

Ninety percent of the people need help, you know, in buying a home or selling one. And I’ve watched it operate all my life and been involved with lots of people who have been in the business and bought — there was a decision in court, and I told Greg to handle it whatever way that seemed best to him. And I’m quite — I’m perfectly happy with the way we’ve handled it, I think.

I’m surprised at the decision, but we get surprised in decisions in the insurance business lots of times. I mean, you know, just think of the various things we’ve faced. And, you know, when 9/11 came along, you know, we never thought something like that could happen, but it happened. And then we didn’t know what was one event or more events.

【批注】政治上的倾向性，起点是照顾弱势群体，然后一层层往外推，推到弱势群体都是最正确的，李光耀说平庸是财大气粗者的奢侈品，巴菲特在这里关注的是趋势，但他最终是乐观的，"我们还是会继续做那些合理、理性的事情"。

17、什么是“大规模普及”？

【27. Berkshire isn't trying to predict winners in controlling climate change】

we will — I hope you’re right. And massive adoption has been sort of a moving target so far. But I hope we get there. 

in effect, the United States, particularly, has been the one that caused the problem the most, and then we’re asking poorer societies to say, well, you’ve got to change the way you live because we lived the way we did. 

【批注】按照汽车保有量的统计口径：电动车在中国有3689万辆，占比10%；美国有570万辆，占比2%。过去的电视机有过"大规模普及"，但同样是"大规模普及"，电视机的"大规模普及"指的是70-80%以上，电动车的"大规模普及"指的是2-10%。

当这个观众说"我们所处的世界已经在变化，而且变得不一样了"，他就知道该说些什么了，这个回答对于提问者很难有什么样的帮助，本来就非常难，但还是做了两个思维上有助于反向校正的例子，出于善意，顺便练练自己的脑子。

18、慈善捐赠：越简单、越正确

【29. Shareholder Ruth Gottesman donates $1 billion to make Bronx medical school tuition-free】

You know, there’s no ego involved, you know, nothing. She just decided that she’d rather have a hundred-plus — closer to 150, eventually — students be able to start out debt-free and proceed in life.

【批注】不想自我炫耀、帮助别人免除债务负担，越简单、越正确。

20、巴菲特关心的金融风险

【31. Buffett would make Abel, when he is CEO, responsible for deciding both which businesses and which stocks to buy】

And we do not want to try and have, you know, 200 people around that are managing a billion each. It just doesn’t work. And I think that when you’re handling the sums that we will have, you’ve got to think very strategically about how to do very big things.

but as the world gets more sophisticated, complicated, and intertwined, more can go wrong.

That is for sure. And the more intertwined and sophisticated the world financial situation gets, the more vulnerable it gets, in a certain sense. It solves a lot of small problems, but it leaves it more vulnerable to large problems.

【批注】

想法在对的方向上都是收敛的，弄"200 个人、每人各管 10 亿美元"在方向上就是错的。

2、"世界变得越来越精密、越来越复杂、彼此联系越来越紧密"

这个判断有点疑问。

（1）大型科技企业的比重越来越大，大型企业和小企业两极化，商业世界的底层不是更复杂了，而是更简单了，巴菲特的话原来指向的是金融衍生品，但问题是底层的企业更简单了，复杂的金融结构卖给谁去？

（2）金融不再像以前那样占据注意力的中心，现在明显更偏向科技，注意力少了、复杂的金融结构会跟着减少。

（3）最大的风险可能不是多级、多层嵌套、会连锁传导的金融风险，少数大型科技企业的风险更大，比如，Amazon、Google、Microsoft的云服务；Apple、Google控制的手机；Nivdia的芯片，这些都集中在少数企业，并且都假设不会有问题。

21、每10年掉出前20名：根本原因是"脑子进水"

【31. Buffett would make Abel, when he is CEO, responsible for deciding both which businesses and which stocks to buy】

And if you look back, as we did a few meetings ago, as to the top 20 companies in the world at 10-year intervals, you realize the game isn’t quite as easy as it looks. But getting a decent result actually is — reasonably — should be reasonably easy, if you just don’t get talked out of doing what (laughs) has — works in the past. And don’t get carried away with fads and don’t listen to people who have different interests in mind than the interest of our shareholders.

【批注】这一句像是提醒GREG ABEL的，弄个平均数是简单的，买个S&P500就有一个体面的结果，但如果预期过高、或者其他形式的噪音侵蚀了理性就很麻烦，每隔10年掉出名单的都是因为这个原因，这个问题很容易想明白。

第一，作为一个思想实验，假设 500 个人一起往前跑，起始数值从 1 到 500，整体速度是 10%；第 1 名始终按 10% 稳定复利前进，其余 499 名整体的速度也是 10%，但内部以随机的方式分配这 10%。用这个设定做模拟，第 1 名第 1 年掉出前 10 的概率只有约 0.04%-0.07%，第 2 年约 0.03%，第 3 年以后接近于 0。这个结果至少说明：只要头部还能稳定复利，单靠后面的随机扰动，其实很难把它掀翻。

第二，换个角度看，其余 499 个主体要超过第 1 名，靠的不是某一年跑得更快，而是要在足够长的时间里持续高于 10% 的复合增速，这是很难的。随机的假设有点夸张但如果放在足够长的时间里，大部分企业的成功是不是有很大的偶然性？时间长了都很难，时不时遭遇的压力、过高的预期都会侵蚀理性，如果每年有1次重大决策，10年10次，30年30次，10次、30次都做对的概率是不是接近于随机？

结论：掉出名单的根本原因不是竞争，而是脑子进水了，如果要进一步深入分析，大部分的进水是因为局部泛化，局部泛化在遭遇变化时更容易产生幻觉，比如，过高的预期，所有的预期都是部分事实、部分幻觉，过高的、超过平均数的预期通常拔高的是幻觉，相比于事实，幻觉更容易拔高。

22、生意越一般，人越重要

【32. Distribution companies don't have a 'wonderful' business, but Berkshire will look at them】

 if we do it, we’ll do it well. We’ll do it right. They’ll have made the right decision. If they don’t, you know, we will find something else to do with the money in the end. And we can always buy a little more of TTI for you, the shareholders, by just buying in our stock, too.

【批注】（1）生意越一般，人越重要；（2）如果要做，就要努力做好；（3）手里已经有很好的东西，再不行可以回购自己公司的股票，急吼吼的追逐就是个白痴，但为什么这样的逻辑只有巴菲特能讲清楚？

23、真正理解了生意不需要关心宏观

【33. We haven't seen anything that makes sense】

So, I wouldn’t — I don’t really think of whether it’s similar to 1999, because I’m not that good on chronology anyway unless something really dramatic happened at the time. I mean, I remember things from 2008 and ’09 much better than I remember whether something happened in 2015 or 1987 or — well, 1987 I remember because of October 19th. But I just don’t think — I don’t think that way. I just look at what I can do every day.

【批注】真正理解了生意确实不需要关心这些东西。

24、Apple：消费者行为与"模糊正确"的量化

【34. 'Seeing something that turns on the lightbulb'】

And in that case, I watched what was happening at Nebraska Furniture Mart. People would leave the store if we didn’t carry Apple, even though we were selling Apple at a price where we weren’t even making any money. It was just so popular that if we didn’t have it, people would leave and go to Best Buy or someplace。And if you know the Blumkins, they can’t stand anybody leaving the store. So they behaved accordingly.

【批注】巴菲特很擅长"模糊正确"的量化，以下是我自己的推理。

1、饮料、手机都是近似刚需（每个人都需要）的产品，这类需求更容易出现商品化的商品，以成本优势取胜；

2、Coca-Cola、iPhone都在近似刚需的地方建立了品牌；

3、Apple更进一步，是一个占据20%份额的奢侈品品牌，可以想一想Coca-Cola/Pepsi，娃哈哈/农夫山泉，这些都有强大的品牌但提价能力要弱得多。

25、便宜不能消除结构性劣势：Paramount的教训

【35. Buffett: I was '100% responsible' for buying Paramount stock and 'we lost quite a bit of money'】

And that happens in this business, too. But, actually, owning Paramount made me think even further. I like to think deeper, but I certainly have — not harder even — about the whole question of what people do with their leisure time.

And, you know, what the governing principles are of running an entertainment business of any sort, whether it’s sports or movies or whatever it might be. And I think I’m smarter now than I was a couple years ago. But I also think I’m poorer because I acquired the knowledge in the manner I did. 

【批注】2022年Q1建仓的股票，建仓成本26亿美元，买入时股价已从高位跌了14%，PE大概是10倍，2023年卖了一部分仓位，2024年Q1全部卖掉，估计亏了15亿美元。

便宜能降低风险，但便宜不能消除结构性的劣势，便宜弄不好也会侵蚀理性，这些卖内容的公司从一开始就不好搞。

（1）明星赚了钱，公司不赚钱，然后有了Disney；

（2）出现了搜索引擎，用搜索引擎在广告上的效率支持内容的产出，比如，YouTube，YouTube一开始就需要大容量的带宽，如果不是Google很难搞定，从硬件基础到后面的内容生产都因为Google的广告业务；

（3）出现了推荐引擎，广告变得更有效率，这就是ByteDance；

（4）出现了人工智能的大模型，Google Veo、Seedance、Sora把内容创作的门槛降到了极致，明星这个高能耗的环节正在承受很大的压力，谁还用明星？用AI直接生成一个的成本要低的多。

Google和ByteDance在竞争中既有广告上的优势，又有内容创作上的优势，Paramount两个都没有，存在结构性的劣势，要生存下来太难了，即使Netflix、Disney都很难。

26、BNSF的估值逻辑：下行保护+上行收益

【36. BNSF is trying to reduce costs to better compete with rival railroads】

we want everybody to have the idea that there’s a lot to be done with every business. You know? So, I mean — it is — you know — we — you know — a fellow named Pete Kiewit built a remarkable company [Kiewit Corporation] in Omaha — building company. Really remarkable. And it is — the question after everything they did, that when something was done particularly well, they’ll — digging a tunnel under the East River or something when they said it couldn’t be done — he would say it would be — he was pleased but not satisfied. And that is exactly the way we want the attitude to be at Berkshire forever.

So it has a net benefit to us from making the same amount of money owning one of the other railroads by owning all of the railroads. And we got 35 billion out during a recessionary period. I think that was the worst quarter, the third quarter of 2009, maybe that the rails had had for a long time. So it’s worked out. 

 if you shut down the railroads in the country, it would be incredible, the effects. But — and it would be impossible to construct now.

【批注】讲了企业估值的基本逻辑。

1、第1个问题是下行保护

依次的顺序是：（1）没有致命的风险；（2）不亏钱；（3）有合理、够得上平均数的利润率，BNSF不存在致命的风险，整个经济系统少不了铁路，重建成本高到无利可图，巴菲特以合理的建仓成本让BNSF有了一个合理的利润率。

BNSF的买入成本是350亿美元，2010年的税前利润是39.88亿美元，当年的资本支出22.74亿美元，折旧15.32亿美元，资本支出超过折旧的部分是7.4亿美元，巴菲特起步要求是不超过15倍的PE，大概是7%，BNSF可能还稍微高一些；到2024年，BNSF的经营现金流76.25亿，投资现金流流出38.39亿，剩余38.39亿，当年分给母公司40亿美元，基本上全部回流给了母公司。

2、第2个问题是上行收益

巴菲特在这个部分用了Kiewit Corporation的例子，能做好要尽力做好，能改进的地方要继续改进。

27、深度低估的存在，本身就是最重要的事实

【37. You can find deeply undervalued securities, but you have to love the game, not the money】

So all I know is the human brain is complicated. But it does — it’s best when you find out what your brain is really suited for, and then you just pound the hell out of it from that point.

【批注】有一个没有说到的点，为什么会有一些深度低估的股票？显然是因为没有人看到本来能看明白的东西，这是一个重要的事实，间接证明了真正能看明白的人很罕见，罕见和如何做到之间存在很大的空间，巴菲特在最后把话收回来了，因为不管是什么，付诸热情去寻找是唯一可行的路径。

28、防范噪音干扰的三个办法

【38. 'You have to be alert to what human nature does to both other people and to you'】

A, you want to be regarded as an asset to the country, because you’ll find more solutions if you are an asset. You owe it to the country anyway. But beyond that, you’ll find more solutions than if you’re regarded as evil or something, and worse yet if you deserve it. 

Which — you know — in a sense, you don’t want to become cynical about life. But almost everybody that approaches you, if you have tons of resources, has got some interest in figuring out how to use your resources to their advantage.

But you have to be alert to how — what human nature does to both other people and to you. 

【批注】可以理解为如何防范噪音的干扰？

1、让自己有价值，成为国家的一项资产，如果你是一项资产，你就会找到更多解决办法；

2、如果你有价值，肯定会吸引噪音，用不着"愤世嫉俗"，这是事实的一部分；

3、不要在遭受噪音的过程中，被人性的负面拖下去，要保持时刻的警觉。

29、选择正确的榜样：Basic trust的稀缺性

【39. How Buffett's parents reacted when he ran away from home】

And I walked in the door in Washington and my mother said, “How come you came back so soon?” And my father said — (laughs) — he said, “I know you can do better.” (Laughter) And I just paid more attention to my father than my mother. So you want to have the right heroes. And you don’t have to have them — it’s not the heroes based on what they’ve accomplished. It’s — you know — it’s — It’s the people that you want to be yourself. And if you copy the right people, you’re off to a great start.

【批注】Basic trust/Basic mistrust，现实世界中存在着大量的Basic mistrust，而Basic trust比较罕见，不能因为Basic mistrust多到不行都选择Basic mistrust，但是大部分人没有真正体会过Basic trust也是事实，没有体会过又认为要朝着这个方向做，并且阻力重重中还必须非常坚定。

30、Pilot交易：对人性受力点非常熟悉

【41. Pilot and new CEO Adam Wright are 'only in America' stories】

Well, I'll make two comments on that. A couple of the directors had their doubts about going in. And in any event, Pilot is working out well for us. And as my friend Sam Butler one time said to me — and he was talking in general about certain kinds of situations — but he said, "Well, Warren, all's well that ends." And that's where we are.

So we’ll go to station five. (Laughter) Well, while we’re getting to station five, I’ll tell you a little bit more about the fellow [Adam Wright] that is now running Pilot, and who you may have met here, that Greg had known for a long, long time. And he grew up in Omaha. And came from a poor family and was raised by his mother.

And came away from this football team. And you’d think, well, another football player, you know. He goes out — and there may be some intermediate parts in the story a little bit — but he buys a gas station and he turns it into something that is huge. So it’s — we’re really delighted with it. And, you know — that’s another kind of “only in America” story. And, you know, how many of us can become an All-American, a number one-ranked team, let alone start a business that (laughter) goes on to these sort of heights? So we feel very good about it.

【批注】对人性的受力点非常熟。

（1）被提及跟Jim Haslam家族的官司，结论是"结局好就好"；

（2）按住提问，为Pilot的交易补充信息，自己这里找了一个非常能干的人：Adam Wright；

（3）表扬Jim Haslam，事情总是有好的一面也有坏的一面，要朝着好的一面观察这个世界。

31、保持头脑敏锐的秘诀：有一个对齐的脑子

【42. 'Anybody that says I did it all myself is just - delusional'】

we’ve got all the heroes from American history and all the wonderful things they did, but how could they say, “All men are created equal,” and then write a Constitution (laughs) that women, you know, women not to be able to own property, and depending on the state, I mean, just terrible conditions.

But anyway, that’s how you learn about what the humans can do. (Applause) And if you — and you’ve got to feel better about the future for your kids than you would’ve felt a hundred years ago, no matter, you know, what the situation is.

【批注】问：保持头脑敏锐、判断出众、身体状态良好的秘诀是什么？

回答分了几步：1、需要运气；2、为什么需要运气，这是一个处处不对齐的世界；3、现在比过去好多了。

想法上的对齐对于每个人来说都很困难，人类的整个历史也表明这个事很困难。AI关于对齐（alignment）的思路是："同一套压缩出来的原则，对"结构相同但表述不同"的两个句子，能不能给出同样的答案？"

以下是巴菲特经常举的例子：

（1）如果一辈子要吃汉堡，不是养牛人，你希望牛肉价格高还是低？

（2）未来很多年你是股票的净买方，你应该希望股市更高还是更低？

"结构相同但表述不同"的两个句子会给出什么样的答案？如果能给出一样的答案，那么这个人的脑子大概率能长期保持敏锐、身体状态良好；如果不能，处处不对齐，那就有他该有的样子。

32、Loser的核心问题是心理上的防御

【43. 'You really get some strange things revealed in a will'】

he actually liked to use codicils, because I think he had, like, 25 of them. And it was kind of his way of writing, well, I’m taking you out of the will because, you know? And so he sort of delighted in explaining through his will what — how he felt about (laughs) all these people.

His opening line, in effect, said, I’m writing this will while I am riding in the economy section of Eastern Airlines number such and such. I mean he believed in getting right to the point of what — (laughs) — what the people who were recipients, how they should live. (Laughs) And he was going to be judging them.

【批注】Loser的核心问题是心理上的防御，说教只会增加对方的防御能力。

33、企业估值：当下的确定性+未来的变数

【43. 'You really get some strange things revealed in a will'】

I mean, I had a few goals when I was 30 or 40, and may have written them into wills then, in terms of what the world needed done, and how the money could be used. And unfortunately, I decided that it just — they just weren’t feasible to accomplish.

But human beings are human beings. And we all have weaknesses and peculiarities and everything else. And don’t be too hard on yourself because you have some of those. But don’t be totally forgiving either. (Laughs) You can change the future. You can’t change the past, but you can change the future.

【批注】企业估值的基本思想：当下的确定性+未来的变数，即使"当下的确定性"抓的很到位，未来仍然充满了变数。

34、AI：几个值得推演的问题

【44. Too early to predict how AI will affect Berkshire and the world】

obviously, you know, anything that’s labor intensive and that — it can create an enormous amount of leisure time. Now, what the world does with leisure time is another question. Whether more leisure time — I know an awful lot of people think when they go to work at first that what they want is leisure time. And what I like is actually having more problems to solve.

But AI is profound. I mean, that’s what makes it — makes it a genie. You know, is what can happen. I’ll — I could tell a few genie jokes, but I better not.

【批注】更多的闲暇时间，填满闲暇时间的可以是社交、娱乐，也可以是相互制造的冲突和矛盾。整个世界的分辨率会因为AI大幅度的提升，可以看到大量原来看不到的问题和需求。

1、这些新增的问题还是不是人类的问题？

2、AI目前还不擅长"模糊的正确"，同样困扰于碎片化，在知识的收敛上受到碎片化的阻力。

这两个问题值得进一步推演。

（1）Basic trust/Basic mistrust，不管是怎么演化的，现实是Basic trust在人群中所占的比例非常少，周围挤满了Basic mistrust，都是前半句的逻辑跟后半句的逻辑对不上、因为和所以之间没有关联的人，这对Basic trust来说是一种对抗的强度，需要在大量噪音中提炼出简洁、有效的知识结构，从进化的角度观察，类似巴菲特的能力显然跟这种对抗的强度有关。

（2）噪音越强、对抗噪音的能力跟着增强，哪天失去了跟大量噪音竞争的环境，智能水平会不会跟着下降？

（3）对抗噪音的原始推动力是什么？AI还在寻找各种各样的激励函数，但是从人类的角度观察，Basic trust/Basic mistrust非常简单，Basic mistrust从根本上不相信"有正确答案存在"，而Basic trust对世界有基本的信任，从来不怀疑"正确答案是存在的"，从一开始就对反方向的信息有生理上的敏感，心理上的怀疑和抵触；

（4）AI能不能像人类当中的Basic trust，从一开始就专注于对抗不信任和噪音，Ilya Sutskever的想法是造出一个超级聪明的15岁少年，如果真能实现，相当于人群中都是15岁的巴菲特和乔布斯，大量噪音跟着消失，这些15岁的巴菲特和乔布斯还能不能正常成长？

（5）如果最后发现，真正的智慧来自跟噪音的对抗，是不是未来某个时候需要在系统中重新投放噪音？最后AI也不见得比人类的进化更有效率，或者两者差不多。

35、财政赤字与心理上的锚点

【45. Buffett more worried about fiscal deficits than quantity of U.S. debt】

And then — It won’t be the quantity. It’ll be whether, in any way, inflation would get let loose in a way that really threatened the whole world economic situation.

But it is — I don’t worry about the quantity. I worry about the fiscal deficit. 

【批注】（1）赚100块，花100块，最保守、最没问题的做法；（2）赚100块，花103块，预计来年能赚103块，这是事业稳定并有所增长的状态；（3）赚100块，花107块，预期能赚的更多（超过3%）但长时间没有实现，时间久了会把幻觉当成现实。

1965年以前的美国：

1、二战后国防开支从1945年占GDP的37%骤降至1950年的不到5%，这一大幅削减与强劲的经济增长叠加，使债务占GDP比率迅速降至可控水平。

2、战后婴儿潮带来的人口红利——大量年轻劳动力进入市场，税基快速扩张，政府收入自然增长，不需要借债就能维持支出。

3、1960年时，社会保障支出占联邦总支出不到13%，Medicare和Medicaid还根本不存在。福利体系的包袱极轻，支出结构简单。

1965年发生的两件事成了转折点：（1）Great Society，这是一个永久性的、自我扩张的福利承诺——Medicare、Medicaid、各类社会项目一旦建立，就有自己的政治生命，极难削减，而且随着人口老龄化会自动变大；（2）越战。

有意思的是当年的财政赤字放在今天并不大，但最终把利率推到1x%以上，可能跟心理上的锚点有关，比如，二战以后的经济快速发展；以前的美元是跟黄金挂钩的，这些锚点跟1965年以后的情况产生了比较大的对比，1980年以前的中国，有谁见过穿超短裙的女性，激动的不行，放在现在看到没有穿的反倒不太习惯。

36、善良是对的，但标签的作用是有限的

【46. 'Be kind - and the world's better off'】

I feel — I don’t think there’s any room in beating up yourself over what’s happened in the past. You know, it’s happened, and you get to live the rest of the life. 

And then I think the one thing that you can aspire to be — because this can be done by anybody and it’s amazing — doesn’t have anything to do with money — but you can be kind. 

And in the end, aspire to be more — or I’m sure many of you are yourself — but just aspire to be more so.

【批注】“Attention Is All”，过去的已经过去，要把有限的注意力放在还有可能调整的未来，这跟企业估值的逻辑一模一样，知道什么重要什么不重要，什么是重要并且可知的，什么是重要但不可知的，Basic mistrust永远不清楚什么是正确的方向，事实也是因为注意力的方向坏了才导致的Basic mistrust，反正校正的难度极大，因为每一个想法的注意力方向都是错的。

巴菲特说的“做一个善良的人”，这是一个心理上的锚点，Basic mistrust更需要这样的锚点，但也可能极端化，起点是照顾弱势群体，一步步往前推，最后认为弱势群体的一切都是对的，李光耀说平庸是“财大气粗”者的奢侈品，这是美国的加州和纽约正在发生的事情。

“做一个善良的人”肯定是对的，有意思的是为什么本来就应该存在边界会变得模糊不清？所以，真正的问题还在于对边界的感知能力，而不是标签，标签的作用是有限的。

37、如果你很幸运，让更多人也变得幸运

【47. 'If you're lucky in life, make sure a bunch of other people are lucky, too'】

And I would say if you’re lucky in life, make sure a bunch of other people are lucky, too.

【批注】通过行为、可以作为榜样的行为让其他人对于人性中好的一面有信心。