What We’re Reading (Week Ending 10 December 2023) - 10 Dec 2023
Reading helps us learn about the world and it is a really important aspect of investing. The late Charlie Munger even went so far as to say that “I don’t think you can get to be a really good investor over a broad range without doing a massive amount of reading.” We (the co-founders of Compounder Fund) read widely across a range of topics, including investing, business, technology, and the world in general. We want to regularly share the best articles we’ve come across recently. Here they are (for the week ending 10 December 2023):
1. Charlie Munger – A Conversation with Charlie Munger & John Collison – Charlie Munger & John Collison
John: [00:15:25] So my question is, how do you think about the quality of the business when overarching tech changes are really going to shake it up?
Charlie: [00:15:32] You’ve got to recognize the tech changes do cause some new businesses to flourish and other businesses that looked impregnable to fail. And that’s one of the realities you have to understand.
John: [00:15:44] So you secretly are a tech investor because your reasoning about the effects of tech on Costco or on…
Charlie: [00:15:51] Yes, it’s just that — take, for instance, pharmaceuticals. The American pharmaceutical industry is better than any other pharmaceutical industry in the whole world. And number two is not remotely even close. So we have one of the great achievements in the whole history of the world in science and technology and so forth. At the same time, there’s a fair amount of sleaze in the way pharmaceuticals are distributed. Everybody rooks the government…
John: [00:16:16] The PBMs, yes.
Charlie: [00:16:17] Yes. And that’s just the system. By and large, we haven’t invested in pharmaceuticals because we’ve got no edge. I don’t know enough about biology and medicine and chemistry to have any edge in guessing which new pharmaceutical attempt is likely to succeed and other people who know those things, not that they have perfect knowledge, but it’s way better than mine. Why in the hell would I play against other people in a game where they’re much better at it than I am when I’m playing for something desperately important to me like a way of feeding my family. So of course, we didn’t go near it.
I would argue that they’re — in practical life, you want to succeed, you got to do two things. You got to have a certain amount of confidence. And you have to know what you know and what you don’t know. You have to know the edge of your competency. And if you know the edge of your competency, you’re a much safer thinker and a much safer investor than you are if you don’t know it. And I constantly meet people, better to have an IQ of 160 and think it’s 150 than an IQ of 160 and think it’s 200. That guy is going to kill you because he doesn’t know the edge of his own competence and he thinks he knows everything.
Partly, Warren and I, we pretty much know what we know and what we don’t know, what we’re good at, what we’re not good at. And one of the things we’re not good at is guessing which new pharmaceuticals. So we don’t even look at it. After all, it’s a big universe out there and if we have to leave a certain kind of investment behind because we lack the capacity to deal with it as well as some other people. That’s all right. We don’t need an infinite number of opportunities…
…John: [00:21:36] So what examples do you prefer of businesses driving capital efficiency without squeezing small suppliers?
Charlie: [00:21:43] Well, Costco is one.
John: [00:21:44] By turning the inventory quickly?
Charlie: [00:21:46] Yes and doing that because they have fewer stocking units and they’re way more efficient.
John: [00:21:51] You’re on the board, right?
Charlie: [00:21:53] Yes. I am somewhat the older member. But Costco, it’s an amazing culture. The whole damn culture of the place is so subtle and it just marches from triumph to triumph. It was smart to have a small number of stocking units flowing through with enormous speed. It was right to have a membership system.
There are three things that Costco didn’t want. Didn’t want people who stole merchandise. They didn’t want the people who used bad checks, and it didn’t want people cluttering up his goddamn parking lot without spending a hell of a lot of money in stores. So a membership system, where they accept only a certain kind of a member, all of a sudden now they’ve got nothing but people who buy a lot per trip.
Costco has always had the lowest shrink rate in the world. Tricks the inside too. So the net theft rate at Costco was always below 2/10 of 1%. That’s unheard of.
John: [00:22:46] I hadn’t thought of the parking lot efficiency with the membership system.
Charlie: [00:22:49] You can’t go to Costco just to buy bottle of iodine, just drop in. You got to be a member and then you got to pay enough, so to an ordinary person, they’re not going to pay an extra $100 to buy a bottle of iodine or something. We keep the peach pickers, the little buyers out.
Sol Price used to say “A business should be careful in the business it deliberately does without”. Of course, that’s straight out of a Munger book. You figure out what you want to avoid. And they want to avoid theft losses, embezzlements, bad checks and cluttering up the parking lot without buying much. And their system caused all those effects at once.
John: [00:23:27] It’s like your first speech in the book, start with the business you don’t want, work backwards.
Charlie: [00:23:30] I know, but it’s so simple.
John: [00:23:32] Any others? Of businesses driving capital efficiency without squeezing suppliers.
Charlie: [00:23:35] There are lots of others. Practically all of the aerospace businesses have learned to make very high returns on capital.
John: [00:23:42] How do they do it?
Charlie: [00:23:43] They specialize in being good at something and handling the government well as a paying customer.
John: [00:23:49] Did you ever look at TransDigm?
Charlie: [00:23:51] Sure. I don’t like that way of making money.
John: [00:23:54] Because the price increases.
Charlie: [00:23:56] It’s too brutal. They figure out something that has a little monopoly due to the defense department regulations, and they raise the price 10 times. And they’re famous for it. I regard that as immoral…
…John: [00:24:25] One of the things you raised in the book is this question of when you have a small number of players in the industry, say, two or three players in the industry, it is not always easy to predict who will earn good profits and who will not. And so the airlines lost money since the Wright brothers versus the cereal manufacturers, very durably profitable. If you’re looking at the business today and you know that the industry will consist of two or three players, how can you predict will those players make money?
Charlie: [00:24:50] I don’t think it’s possible to be 100% accurate in making these predictions. But certainly, we’re looking backward, the people who had branded profits like coffee and oatmeal and so forth, made very high profits and airlines basically made no profits at all for their shareholders.
John: [00:25:09] But the airlines were branded goods.
Charlie: [00:25:11] But everybody had big capital equipment if they didn’t use it, they obviously were losing a lot of money. So that everybody was almost forced into a very destructive competition by the logic of the individual situations. There are a lot of businesses that are very hard to make money in permanently.
If you want to go into the business like restaurants, most people fail, small percentage of restaurants even last long enough to make a living to the people who own them. Too competitive. That’s why they fail. Just like there are too many deer on an island and no predators, pretty soon there are too many deer. So all the deer suffer because there are too many of them.
John: [00:25:49] But again, to go back to this question, if I want to understand, will the business be like an airline or like a cereal company. Is this then the ongoing capital expenditure that’s required where airlines fundamentally, they have lots of CapEx on an ongoing basis, it’s like the original Berkshire textile mills?
Charlie: [00:26:02] The airlines are like a guy who builds a big hotel and it’s just sitting there and he makes some incremental profit from filling it. And if it’s got up and staff, that’s better than just letting it sit there vacant. It almost forces irrational intense competition. The same thing does not happen within cereals.
John: [00:26:19] BNSF was one of the biggest acquisitions you guys did. And my sense from the outside is that it’s maybe even been more successful than you would have expected. Is that accurate? Or did you expect it to be this successful?
Charlie: [00:26:31] The railroads were a lousy investment. There were a few people when they were first created that basically stole all the money by milking the government, bribing legislators and doing all kinds of terrible stuff. But by and large, most railroads are lousy investors, like the airlines for a long time. And finally, it got down after years of fighting unions and consolidations, so you get down to a few big systems.
Now there are just two big transcontinental systems, and we’ve got one of them. Of course, that’s a less competitive market than it was — than it existed earlier when there were 100 different railroads. But early railroads when they were terribly competitive, they were terrible places to invest money.
John: [00:27:12] But again, airlines, bad business, not a good investment.
Charlie: [00:27:16] Early railroads, bad business.
John: [00:27:19] Early railroads, yes. Railroads today still require a lot of ongoing CapEx.
Charlie: [00:27:23] Yes, but they’re so dominant. Once you have a railroad that can put shipping containers on that stack too high on tracks. It’s one of the most efficient ways of transferring assets all over the country. It’s way more efficient than trucking. So that they have a system that just accidentally happened. Nobody anticipated you’d be able to double the capacity of the railroad just by shoving containers one on top of the other. So they got very efficient finally. And now they’re so efficient. They’re more efficient than anything else. And of course, they do well…
…John: [00:32:38] You have been famous for criticizing gold earlier on and now cryptocurrency.
Charlie: [00:32:46] I like to gold a lot better than I like cryptocurrency.
John: [00:32:49] You’ve criticized both.
Charlie: [00:32:51] Before there was cryptocurrency, I never bought gold. So I didn’t like gold. But I don’t hate gold as an investment as much as I hate cryptocurrency. I think cryptocurrency ought to have been driven out as illegal.
John: [00:33:03] At the risk of maybe getting ejected from the premises, if I can try to defend cryptocurrency, isn’t the perspective you have where — I think you would say, invest in a productive business. Isn’t that a reasonably U.S.-centric perspective, where absolutely, we have a great currency here. We have a great respect for property rights here. If you’re in Turkey and their property rights aren’t as strong, the currency is inflating 80% a year as it has this year, then the ability to move your wealth…
Charlie: [00:33:34] Well, if I lived in Turkey, I might do something odd. Buy my gold if I were in Turkey, but I would never buy cryptocurrency.
John: [00:33:39] Even in Turkey?
Charlie: [00:33:40] No. I don’t think that buying a percentage of nothing is a good investment, even though it’s hard to create more nothing.
John: [00:33:47] But isn’t gold functionally an investment in the percentage of nothing?
Charlie: [00:33:50] It is similar, except it’s been established so long as a…
John: [00:33:57] An agreed upon store of wealth…
Charlie: [00:33:58] With the history we have and with the need for a currency and a currency that is backed by something and gold is hard enough to mine and so forth. Gold is a perfectly reasonable thing to use as a currency. And the evolution of use of gold as a currency was a very good thing for civilization. I don’t have the feeling that gold is evil. Gold helps civilization develop. But I think cryptocurrency is scumball activity. And I think by and large, the people who promoted it are scumballs or delusionary. And I don’t know which is worse, being a scumball or a delusionary. But I think they’re both pretty bad.
John: [00:34:31] Some people can manage to be both. There’s plenty of scams in crypto. That’s absolutely not up for debate. But are we talking about questions of degree here between gold and cryptocurrency, where they are societally agreed upon stores of value, which trade above…
Charlie: [00:34:46] Let’s put it this way. If we didn’t have gold, we might have invented something like cryptocurrency as a substitute. But once we have gold and fiat currencies that are now long established, we don’t need to add in cryptocurrency.
John: [00:34:59] But isn’t cryptocurrency handier? If you can work with it in just software, you don’t need to actually go get some physical gold, trade it, melt it down. It’s much harder to seize cryptocurrency than it is to seize gold in an autocratic regime.
Charlie: [00:35:10] You don’t have to bother with any physical inventories or anything has any intrinsic value. You can create system very efficiently dealing in it. I don’t want to officially deal in nothing and craziness. I want to make it illegal. All nations have had anti-counterfeiting laws. And I think the anti-counterfeiting laws ought to have been used to totally bar cryptocurrency.
John: [00:35:33] But nothing’s being counterfeit here.
Charlie: [00:35:34] Well, if I am a nation and I have a currency, I don’t want a new currency established.
John: [00:35:38] But it’s not really a new currency. It’s a new store of wealth.
Charlie: [00:35:42] You can call it a store of wealth, I call it a store of delusion. I don’t think it’s good to participate in delusion even when it gets quite common. A second medium of exchange widely used. It’s ideal for drug dealers, dope dealers, scam artists of various kinds. Every kind of criminal you can imagine. Very good in extortion, kidnapping.
Why would we want a wonderful crime facilitating new medium of exchange? Why wouldn’t we just say this is like counterfeiting? You’re coming into the government’s business and you’re trying to create a fiat currency and you can’t do that. It’s a feel they don’t…
John: [00:36:18] All right. Well, I will agree to disagree on crypto. But…
Charlie: [00:36:20] You don’t have to agree. I can handle it if you like crypto. I don’t like it, but I can handle it.
John: [00:36:25] We’re staring into a recession, potential stagflation. What advice do you have for people thinking about how to work their way through the…
Charlie: [00:36:35] I have one standard set of advice for all difficulties, suck it in and cope. That’s all any human being can do, suck it in and cope. Partly, you have to be shrewd. That’s one way of coping is to be as shrewd as you can possibly be. But that’s my recipe. And I must say it’s worked pretty well for me. It will work very well for any other person who uses my methods…
…John: [00:37:59] How do you feel about American society over the coming decades?
Charlie: [00:38:03] Old men have always tended to think that new generation is going to hell. The old Romans, o tempora, o mores. That goes back to the earliest civilizations we had. The old guys were saying when I get out of the world, it’s going to hell. And it really wasn’t going to hell net by and large. But I do not like the way politics has morphed in my lifetime in the United States. I don’t like democracy. The way it actually morphed into existence with these primaries and the dominance of two parties where only the most extreme members of each party have a lot of pulling power and therefore, they control the nominees and so forth.
I think our way of getting nominees is deeply flawed now. It may have worked pretty well up until now. It worked better when we had those old, crooked bosses in the cities then it’s working now with the primaries. I wish those old, crooked bosses would come back and replace the present primaries. Wanted to control the patronage so they actually nominated some pretty good people like Teddy Roosevelt. And these modern primary systems, the worst people often win…
…John: [00:39:10] How do you feel about declining birth rates?
Charlie: [00:39:11] It creates a different kind of a world. Well, I don’t see that mankind would be at all smarter if everybody had six children. I think that just jams up population way too much starting with 7 billion for the whole world. So I think it’s good that the population is growing more slowly. But do I think it is good for people to be quite self-centered below 35 and then get married compared to marrying at 21 or 22 and having a lot of children?
No, I think the people who married at 21 or 22 and grew up fast because they had to because they have those young children. In a sense, I think they were a luckier generation than the people who came along with all these different options and who delay marriage into late age and have one or two children, I’m not at all sure it’s good for the people who are having these new options, but it is good for the population…
…John: [00:43:15] Where do you think the world is getting worse?
Charlie: [00:43:17] I think we have a political game problem that’s probably as bad as we’ve ever had. We have some crazy dictators on the verge of creating a nuclear war. We’ve got lots to worry about. The world has never been a perfectly safe place and it isn’t now.
John: [00:43:32] Kind of a societal version of your avoiding mistakes framework from Poor Charlie’s Almanack, where societies need to avoid the major mistakes just like individuals do, avoiding nuclear war.
Charlie: [00:43:42] We’re lucky to have done it so far. But if enough crazy people have enough hydrogen bombs, there will eventually be enough hatred, we’ll have an atomic war of some kind someday. You can almost count on it. So you can say that our generation, it was quite unlikely, but I think it’s getting more likely and not less…
…John: [00:46:15] And what’s an example of where you are more multidisciplinary than the architects in some of the buildings you’ve designed?
Charlie: [00:46:19] If you take the building, the graduate residence at the University of Michigan. They had a magnificent site with a parking lot. They had no other site. They’d used up all the land in the dormitory. They have a second campus, but on their main campus, they’d used up all the sites. And there is one little parking lot left. And I realized that if they used their power of eminent domain and doubled the size of that parking lot, they’d get it with a big square building on the site tha would hold a lot of graduate students.
But there was no way to do that without creating a window shortage in some of the bedrooms. And I also knew that it didn’t matter that there was a window shortage in the bedrooms because I went around Ann Arbour and saw the private builders in Anna Arbour now have already created apartment rooms with no windows and relying on artificial light. And I walked side-by-side exactly identical bedroom, one with a real window and one with just a blank wall. And the one with just a bank wall renting for 10% less.
So it wasn’t much of a problem. I looked for the evidence and then once I realized that, I could do all kinds of wonderful things in that building once I got over this prejudice that it was absolutely required under any and all circumstances that every bedroom have a window. So it’s just an example of just the most elementary common sense. I looked at the evidence at Ann Arbor. I understood geometry well enough to know. And then too, I was well aware that every ship has exactly the same problem. Every ship has a window shortage automatically. Every cruise ship. Yeah, and they pay $20,000 a week to be on the ship and so forth.
And if they don’t want a little light, they walk out of the ship and go into one of the common rooms. And of course, that’s what I arranged they do in the dorm. So I was following correct precedents from marine architecture. But show me an architect that’s learned anything from marine architect. I think you could go into any school of architect in the country and you won’t find anybody studying marine architecture. They think it has nothing to do with it. It has a lot to do with what they’re doing. If you don’t look, you won’t find.
John: [00:48:27] I feel like another example of understanding the customer is giving the students in the dorms single rooms where most people design…
Charlie: [00:48:32] Oh, well, that — talking about insanity. Now, I have sent a lot of children through a lot of graduate education. And I’ve never had a child that liked being in a room with one or two other unrelated people sleeping in the same room…
…John: [00:49:19] Why did this shared delusion persist for so long?
Charlie: [00:49:22] What happened was that the fire codes, they worry that the fireman would need a ladder to go and look through the window and crawl in through the window and haul somebody who had passed out from smoke. So they required that every sleeping space have a window. So the fireman could crawl up on a ladder. There were two things wrong with that.
One, it never happened. Nobody could find a case were a fireman — they would crawl up by ladder and look through and they had found somebody lying in bed passed out of smoke. And two, of course, a modern building with automatic sprinklers, that’s why there was going to be zero.
And that’s why the fire codes changed. And when the fire codes changed because — but the people are used to doing it in a certain way. Of course, they keep doing it the same way they’ve always done. Isn’t it the Mayo Clinic is one of the best places on earth in terms of an admirable culture. They kept doing hip replacements by a procedure that the doctors knew how to do because the new one that was better for the patient was very hard for the doctor to learn. And so they just kept doing it the old way. Architects are no different. They do what they’re used to.
John: [00:50:32] Again, for say, someone who’s 25 or 30, is the lesson that there are a lot of $20 bills lying on the sidewalks? There’s a lot of inefficiency in the world to be rectified that people should not assume the world works efficiently?
Charlie: [00:50:45] Well, of course, there’s always a lot of things that can be improved, always a lot of people who are getting ahead by doing something new. And that’s one of the pleasures of modern civilization. And imagine a postal clerk in the United States can go to Hawaii on a 2-week vacation on a superjet and have a nice time. A postal worker could do that in the world that you’re up in.
You can learn a whole new profession just punching buttons on the Internet and so forth, so the possibilities of self-education is fairly enormous. So all kinds of things have been greatly improved. Of course, that causes new opportunities for some people, and it causes absolute economic destruction from certain people who get obsoleted.
Imagine the Kodak company, which hired all the PhD chemists, totally dominated the chemistry of film and so forth and had the most reliable trademarks in the whole world. Go through Africa when I was young, there are 2 things you always saw: a Coca-Cola and Kodak. That was the brands all over Africa, the poorest villages. And of course, Kodak went totally broke because somebody invented a new way of taking photographs and developing photographs. And it just obsoleted their whole damn business, and Kodak wiped out its common shareholders. That happens all the time, that kind of thing. And you can’t blame the management for it and say, “Well, didn’t Kodak invent its own destruction?” That’s hard to do.
I mean for human nature, you’ve got a business as big as Kodak, everybody’s lived over for years. They’re like the surgeons who didn’t want to learn a new trick that was lot harder to learn when they were old. People don’t welcome having to learn something new. It’s really hard to learn. Everybody would rather get ahead using what he already knows…
…John: [00:53:18] You spend a lot of time in the book talking about businesses that are win-win for both sides and the importance of this for their long term.
Charlie: [00:53:24] How can anything be more important? It isn’t just that it works better in terms of creating plenty for all. It’s better morality. Of course, both sides want both sides to win, that’s more moral than trying to take advantage of other people when it’s so obviously the right way to live and it’s the right way to do business…
…John: [00:55:56] So you wouldn’t invest in drugs, tobacco or the Grateful Dead?
Charlie: [00:55:59] No, that’s correct. I would not. When I sell you a tennis racket for $100, one side gets the tennis racket they’d rather have than a hundred dollars they’re partying. The other guy, he likes what he’s getting, too. It’s win-win.
That’s the beauty of capitalism. It makes win-win transactions very easy and almost automatic. That’s such a hugely important idea. And people like Bernie Sanders and Elizabeth Warren, both of whom I regard as quite talented in some ways, but they just don’t get it.
John: [00:56:27] But I think you mean that as a backhanded compliment.
Charlie: [00:56:29] It’s both a compliment and a criticism.
John: [00:56:32] Is the fundamental thing they don’t grasp that a lot of the win-win businesses are net positive and win-win for both sides and they…
Charlie: [00:56:40] It’s automatic in a capitalist transaction, unless one side is making a big mistake. And most people are pretty good at not making mistakes over and over again with their own money.
John: [00:56:49] It’s not fully automatic, right? We can…
Charlie: [00:56:51] No, it’s not. But a lot of good happens automatically.
John: [00:56:54] Do you worry about the rise of this faction of the political spectrum who don’t really believe in capitalism?
Charlie: [00:57:00] Of course, look at the misery that’s happened to the Russian people. They didn’t like their old system with a bunch of serfs serving a bunch of landlords and so forth, corruption and so forth, so they went to something worse.
They were rebelling against something that was awful, so they substituted something that turned out to be actually worse. It’s hard to create a new form of government worse than Russian serfdom, but Russia has managed to do it. And not only that. They’re proud of having done it. You should never be proud of your defects.
John: [00:57:30] What are Berkshire’s defects?
Charlie: [00:57:31] We haven’t eliminated all mistakes of judgment or even all mistakes of morality. So nobody gets anywhere near perfect ever in human affairs. It’s not exactly a defect. A lot of what worked for us in the early days, we can’t do anymore because the world is more competitive.
The low-hanging fruit has all been picked, and we can’t get fruit out of barren branches where the fruit has gone away. And so we have to go to something else. And of course, that’s harder. A lot of people have that problem, and they go to the new systems in new ways.
John: [00:58:01] I’ve always liked the quote capitalism is how we take care of people we don’t know.
Charlie: [00:58:05] It’s certainly remarkable how it works. I like a social safety net, but I’m different from other people. If I were running the government, the modern civilization, I would be quite liberal at rewarding everything that can’t be faked, like being blind or not blind or something. I’d just give a very blind person a lifetime pension, which goes up with inflation.
If life is tough enough for you, we can afford to do it, and you and your handlers can figure out how you use the money. So I would be very liberal. I would give anybody any education right through college, courtesy of the government, but it would be meritocratic. You have to be able to do the work or you don’t qualify for the benefit.
So I wouldn’t let people pretend to be learning things in some half-assed institution and send the bills to the government. But places like Caltech or MIT, anybody could get in and do the work, if I was the government I’d pay for it all the way through college and graduate school, which they do in places like New Zealand and Australia and so on.
Again, everything in medicine, that is almost automatic, I would pay for that, too. But would I pay for Freudian analysis? No. Stuff that can be gamed and it was crazy, I would not pay for. And I wouldn’t allow the people to get rewards for low back pain, even though they have real low back pain. And it’s easily faked. I wouldn’t pay. It just causes too much cheating and the cheating gets to the eventual and so forth.
I would just say we can’t do that. It’s not that we don’t sympathize with your low back pain and your poor life adjustment? But we can’t give lifetime pay just because you say, “My lower back hurts.” or, “my life adjustment is imperfect.” That’s the way I would organize the government. Nobody thinks the way I do. I feel lonely. I would be quite generous, but I will be quite tough on people with low back pain or psychological problems…
…John: [01:12:23] If Patrick and I came and put you on Stripe, what would you want to understand about the business? What would your concerns be?
Charlie: [01:12:30] That’s an interesting question, considering how much Berkshire Hathaway has made out of other payment systems, including American Express. We recognize the power of having a dominant position in payments in a way that’s very efficient. And of course, anything in modern payments that enables all this Internet stuff is very useful. So you’ve come into a field and made a contribution and made yourself very useful.
I’m for all these payment systems that get better and better. So I think you’ve made your money honorably and you’ve made a lot of it, and good for you. I admire what you people have done. Why wouldn’t I? I regard everything that you’re doing as a little bit threatening to American Express, but American Express actually has a position where it’s like Hermes or something, and so it won’t necessarily be ruined by Stripe.
John: [01:13:22] In evaluating a business like Stripe, what questions would you want to answer for yourself?
Charlie: [01:13:26] Is it likely to remain forever as a money generator? And that’s a more complicated subject. It’s hard to know how the world is going to evolve. If Kodak could suddenly be obsoleted away, maybe it’s not utterly unthinkable if Stripe could.
The company that dominates software for architects, terribly prosperous company, but some other companies come up in that field a lot and it no longer dominates as much as it did. So not everything in software always wins. So I do not have the feeling — the venture capitals tend to think everything in software is always going to win. I don’t believe that for a minute…
…John: [01:14:48] Why has NetJets done so well?
Charlie: [01:14:50] It’s better in its niche than anybody else. In NetJets, the whole culture, safety is first, customer service is second. And after that, we’ll start worrying about the capitalists who own NetJets. And of course, there’s enough fanaticism of that kind of a culture. We create a hell of a product for the person who can afford anything. And ours is better than anybody else in the country, and it’s now big. It’s a big business. And we have yet to kill our first passenger. All these many years, we’ve never killed a passenger…
…John: [01:15:33] I can feature the magazine ads. NetJets- “No one has died yet”.
You’re very bullish on China. Why?
Charlie: [01:15:41] Well, first reason is that their economy was growing faster than ours. That isn’t necessarily true as we consider this exact minute, but for a long time, that economy grew a lot faster than ours. Number two, we could get way better and stronger companies at a much lower price in China than we could get in the United States. Now on the other side, we had to take the political risk of buying into a peculiar system of government that’s not different from ours.
As long as we were getting enough bargains, I was willing to run the — as with part of our assets is we would never invest all of our money in China, for Gods’ sake. But we were certainly willing to invest part of it. That’s perfectly logical. And of course, we were investing through Li Lu, he was a very exceptional money manager. And we put all those 4 things together, the ones, of course, that made sense…
…John: [01:17:19] How does the current geopolitical hawkishness change your view on investing in China, if at all?
Charlie: [01:17:24] Obviously, I’m more uncomfortable now than I was. The guy who changed the whole system and said, “I don’t care if the cat is black or white as long as it catches mice.”, he wanted the goddamn economy of China to work like Singapore’s. Of course we love that guy. And the new guy isn’t quite as much like that guy as we would consider ideal. We think the political risk in China should be run, and I think we should go out of our way to have a lot of friendly relations with big atomic powers.
Both China and the United States ought to get along with one another as a matter of wholly duty because they’re 2 big atomic powers. And the way you get along best is we should carefully work out a bunch of win-win transactions between us and China and actually work to make them work even better. That is the right policy in the United States.
We should not be trying to discipline China by telling them like a nattering nanny how China ought to behave and say, “We know better. We’re a democracy and you’re not.” We have a lot to be ashamed of in our own form of government. We shouldn’t be going around lecturing everybody else. And we should organize win-win transactions with China. Anything else is madness.
And for a long time, we had that. You can argue that China came to modernity primarily in win-win transactions with the United States because we’re so open to their imports. That’s what enabled them to get ahead so fast. And I’m proud of that, and I’m glad we helped them. And I want to do more of it. I don’t want this hostility on both sides.
John: [01:18:53] Tom Wolfe wrote a short story about Bob Noyce. I’m a huge Tom Wolfe fan of his books, but he has a great short story about Bob Noyce. And you can read the short story as it’s really about Grinnell, Iowa and the effect of Midwestern culture in Silicon Valley.
Charlie: [01:19:11] It’s a huge success, of course. And the success is interesting, but I would argue that the failure of Intel was just as interesting a story. Intel was on the ground floor of modern chip making. Absolutely ground zero. They were at the absolute best place. And they just grew and grew and so forth. And they eventually lost all their leadership completely, and they’re just a little pissant company compared to the big guys now.
John: [01:19:40] Why did that happen?
Charlie: [01:19:41] Firstly, some of that’s inevitable. In competitions, somebody are going to lose. It’s — partly it demonstrates the inevitable even if you’re successful, so a little guy that really scrambles, be sure that there’s some accidents, but partly, they were so interested in always reporting more earnings. They didn’t go to the leap enough, just stay on top.
If you’re serving along the edge of a new development like that, you have to just absolutely be going flat out all the time, and you have to be leading all the time. Berkshire, we don’t have to invent new things, particularly, compared to most places. They’re in the business of inventing new things, and you have to be totally fanatic.
And the truth of the matter is that the people in China were way more fanatic than Intel. In China, you had one old guy that controlled the place and he was a fanatic, and Intel had an army of bureaucrats, and they were interested in their executive rewards and the way the price earnings ratios and the approval of Wall Street. A whole lot of other things. And they were powerful. Now they look good for a while just by using their power to make the earnings go up.
But they should have been using their power to make sure their goddamn chips stayed way ahead of everybody else. And they had to be a totally reliable supplier, which they weren’t. They disappointed a lot of customers, and you can’t disappoint customers if you wanted to have a Mayo system of trust. That’s the interesting part of that, not the Noyce story. The story of the failure of Intel was the great story there…
…John: [01:31:17] Is the secret of Berkshire’s culture just the anti-bureaucracy bend? Could you sum it up…
Charlie: [01:31:22] Berkshire is pretty extreme in culture. We are deeply aware of how bureaucracies tend to create their own internal dynamics so that everybody protects everybody else and nobody changes anything, ruffles any feathers. And the net result is that a lot of bureaucracies make some very stupid decisions and we try and avoid that.
But the way we’ve done it, mostly, is by not having anybody around. They can’t be bureaucratic if they’re not there. There is nobody in the head office. So we avoided the bureaucracy. We just don’t want other people to do it. Nobody else is as extreme as we are in that. It’s a huge advantage to us.
And another thing is, we like very trustworthy people. I’d rather have a brief telephone with somebody I trust than I would a 40-page contract prepared by the finest law firm in the world with somebody I don’t trust. And so we like to deal with trustworthy people and to be able to count on their oral promises.
If you look to go into a Mayo operating room is what I call a seamless web of deserved trust. The surgeons trusting the anesthesiologist, the anesthesiologists trusting the surgeon, the nurses are trusting — everybody trusts everybody else. There’s no bureaucracy at all. They don’t have time for bureaucracy.
It’s in patients’ interest to get it over as soon as possible. And so that seamless web of deserved trust can do these very complicated procedures. We like a business system that operates as much as possible like a Mayo operating room, and that requires having very good people who are experienced enough with one another to trust one another.
John: [01:32:55] And that trust is internally between the Berkshire folks or between the Berkshire folks and the managers?
Charlie: [01:33:00] Both. We want the internal and all the Berkshire people to trust one another internally, and we also want the customers to trust us. We’re all for trust. Trust is one of the greatest economic forces on earth.
2. Charlie Munger’s Life Was About Way More Than Money – Jason Zweig
It’s 1931, and a boy and girl, both about seven years old, are playing on a swing set on N. 41st St. in Omaha. A stray dog appears and, without warning, charges. The children try to fight the dog off. Somehow, the boy is unscathed, but the dog bites the girl.
She contracts rabies and, not long after, dies. The boy lives.
His name? Charles Thomas Munger.
Charlie Munger, the brilliant investing billionaire who died on Tuesday in a California hospital 34 days before his 100th birthday, told me that story when I interviewed him last month. I’d asked the vice chairman of Warren Buffett’s Berkshire Hathaway BRK.B -0.75%decrease; red down pointing triangle: What do you think of people who attribute their success solely to their own brilliance and hard work?
“I think that’s nonsense,” Munger snapped, then told his story, which I can’t recall him ever publicly recounting. “That damn dog wasn’t 3 inches from me,” he said. “All my life I’ve wondered: Why did it bite her instead of me? It was sheer luck that I lived and she died.”
He added: “The records of people and companies that are outliers are always a mix of a reasonable amount of intelligence, hard work and a lot of luck.”
I had the extraordinary good luck to get to know Charlie Munger in the past two decades. If you think his life was only about piling up money, think again. Few people have ever been wealthier, in all the senses of the word, than Munger was.
Those who know only a little about him think Munger was a paragon of how to pick stocks—which he was. But those who knew him well consider him a moral exemplar—someone who showed how to think clearly, deal fairly and live fully. He took nothing for granted.
More than almost anyone I’ve ever known, Munger also possessed what philosophers call epistemic humility: a profound sense of how little anyone can know and how important it is to open and change your mind…
…“Part of the reason I’ve been a little more successful than most people is I’m good at destroying my own best-loved ideas,” Munger told the Journal in 2019. “I knew early in life that that would be a useful knack and I’ve honed it all these years, so I’m pleased when I can destroy an idea that I’ve worked very hard on over a long period of time. And most people aren’t.”…
…Munger deliberately kept himself surrounded by people he liked. “Many of the richest people have holes inside of them that they’re always trying to fill,” Munger’s friend Peter Kaufman said last month. “But Charlie knows you can’t fill those holes with money. That’s why he spends so much time with friends and family.”…
…One lesson: the importance of what Munger called “a seamless web of deserved trust” in which a company deals fairly with employees, customers, competitors and other constituencies.
“If you’re structurally adversarial to those adjacent to you in the ecosystem, maybe you prosper for five years,” said Collison, “but not for 75 years!”…
…“You know how a lot of old people say, ‘At my age I don’t even buy green bananas’?” regular guest John Hawkins, co-founder of private-equity firm Generation Partners, said recently. “Well, Charlie is buying green bananas by the truckload. He’s making investments for the next 10, 20, 30 years. He has his foot on the gas and is not taking it off.”…
…He mocked the marketing of short-term investment performance by telling a story about a man who walks into a fishing-tackle store and sees a bunch of gaudy, iridescent lures. “My God, they’re purple and green!” he says to the owner. “Do fish really take these lures?” The store owner answers, “Mister, I don’t sell to fish.”…
…Then I asked what he might want for an epitaph of no more than 10 words.
His reply was immediate and full of epistemic humility: “I tried to be useful.”
Not “I was useful.” That would be for other people to judge. But “I tried.” That much he knew.
3. What Will It Take for China’s GDP to Grow at 4–5 Percent Over the Next Decade? – Michael Pettis
There are two different groups of economists in China that believe that with the right—albeit very different—set of economic policies, China’s economy will be able to grow sustainably by 4–5 percent for many more years. One group argues that China must maintain the investment-driven and manufacturing-intensive strategy it has followed during the past three to four decades. The other group argues instead that China can maintain high growth rates only if it sharply reduces the investment share of GDP and replaces it with a greater reliance on consumption, something which Beijing has been trying to do for over a decade…
…Can China maintain high GDP growth rates driven by high investment? Some simple arithmetic is useful here. Globally, according to the World Bank, investment represents on average 25 percent of each country’s GDP and has remained within a tight range of between 23 percent and 27 percent during this century…
…China, however, is a huge outlier. It currently invests 42–44 percent of its GDP. What’s more… for the past two decades China’s investment share of GDP has never been below 40 percent; it reached as high as 47 percent in 2010 and 2011. In the previous two decades, the investment share of GDP was lower, but it still exceeded 35 percent on average, leaving China during the past four decades with the highest investment share of GDP, and the fastest growth rate in investment, in history.
The obvious implication is that while China accounts for a disproportionately small share of global consumption, it accounts for a disproportionately large share of global investment… According to the World Bank, China’s $18 trillion economy accounts for just under 18 percent of global GDP, making it the world’s second-largest economy after the United States, which accounts for about 25 percent. But China comprises only 13 percent of global consumption and an astonishing 32 percent of global investment…
…if China maintained its high investment share of GDP—in other words, if investment continued to grow as fast as GDP—and GDP grew at rates of 4–5 percent for the next decade, China’s share of global GDP would rise by less than 3 percentage points, to 21 percent, while its share of global investment would rise by more than 5 percentage points, to 38 percent. Its share of global consumption, however, would rise by well under 2 percentage points, to less than 15 percent.
Can China really account for 38 percent of global investment while its economy comprises just 21 percent of global GDP and 15 percent of global consumption? Every $1 of investment has required approximately $3 of consumption globally to sustain it during this century. In China, however, $1 of investment is balanced by only $1.30 of consumption. If the global relationship between consumption and investment held over the next decade, an increase in the Chinese share of global investment from 32 percent today to 38 percent in a decade would require that the rest of the world disinvest to accommodate China’s domestic imbalances.
To give a sense of just how extreme this requirement is, it would mean that to prevent a global overproduction crisis (which would hit China especially hard), the rest of the world would have to agree to reduce the investment share of its GDP by roughly 1 full percentage point, to 19 percent of GDP, well under half of the Chinese level. Needless to say, this is very unlikely, especially with the United States, the EU, and India putting into place policies aimed at boosting domestic investment.
What’s more, to the extent that the surge in China’s debt burden is driven by its extraordinarily high investment share of GDP, it would require China’s debt-to-GDP ratio to rise from just under 300 percent today to at least 450–500 percent in a decade. Given the huge difficulties the Chinese economy is already facing at current debt levels, and the difficulties Beijing has had in its attempts to reduce the debt burden, it is hard to imagine that the economy could tolerate such a substantial increase in debt…
…Globally, according to World Bank data, manufacturing represents 16 percent of GDP and has ranged from 13 percent to 17 percent during this century.
China, once again, is an extreme outlier, with manufacturing representing 28 percent of the country’s GDP. This share had declined from 32 percent in the decade before 2020, but it has risen in the past two years. This recent increase is not surprising. As a consequence of the contraction since 2021 in China’s long-lasting property bubble, there has been a major, policy-driven shift in investment from the property sector to the manufacturing sector, even though the evidence suggests that investment in Chinese manufacturing has been constrained by weak demand—not by scarce capital—so that even more investment in the manufacturing sector implies a further growth in excess capacity (that is, growth in domestic capacity that exceeds growth in domestic demand)…
…While China accounts for 18 percent of global GDP and only 13 percent of global consumption, it currently accounts for an extraordinary 31 percent of global manufacturing. If China maintained annual GDP growth rates of 4–5 percent while also maintaining the role of manufacturing in its economy, its share of global GDP would rise by less than 3 percentage points in a decade, to 21 percent, even as its share of global manufacturing would rise by more than 5 percentage points, to 36 percent…
…To accommodate this and prevent a global overproduction crisis, the rest of the world would have to allow its manufacturing share of GDP to drop between 0.5 and 1.0 percentage points. It would also have to allow a surge in China’s trade surplus—currently equal to nearly 1 percent of the GDP of the rest of the world—as a 5–8-percentage-point increase in China’s share of global manufacturing would be backed by a 2-percentage-point increase in China’s share of global consumption.
Again, this is very unlikely, especially with the United States, the EU, and India enacting policies aimed at protecting and boosting domestic manufacturing. In fact, given China’s determination to increase its reliance on manufacturing to drive growth, I expect global trade relationships to deteriorate sharply in the next few years as the world’s major economies battle over their respective manufacturing sectors…
…The net result would be persistent downward pressure on global demand as major economies competed by subsidizing production at the expense of consumption. This would only worsen global trade relationships because, in the end, only economies that were willing to protect their manufacturing sectors, or maximize the subsidies they delivered to domestic manufacturers, would be able to prevent their manufacturing sectors from contracting as a share of total GDP…
…If they set off a global trade conflict involving the United States, the EU, India, and Japan, the results would be especially painful for countries such as China that rely on large trade surpluses to balance weak domestic demand with an overreliance on manufacturing to drive growth.
That’s because without sustained trade surpluses, there are only two ways a country can balance excess supply with weak domestic demand. One way involves a painful and potentially disruptive collapse in production, as occurred most famously in the United States in the early 1930s, when it had to try to resolve its huge trade surplus in a contracting world economy exacerbated by beggar-thy-neighbor trade and currency policies. The other way is to boost domestic demand as quickly as possible…
…To put it another way, if China wanted to maintain GDP growth rates of 4–5 percent, Beijing would have to engineer policies that caused consumption to grow by at least 6–7 percent a year, with investment growing at roughly 1 percent annually.[3] Any lower consumption growth rate would mean that China could not rebalance its economy in a decade and still maintain current GDP growth rates.
If China pulled this off, at the end of the ten-year period its GDP would comprise 21 percent of global GDP (up from 18 percent in 2022). Its economy would be far more balanced, with investment comprising 29 percent of global investment (down from 31 percent in 2022) and consumption comprising 18 percent of global consumption (up from 13 percent in 2022). In that case, as its share of global GDP would rise by nearly 3 percentage points, its share of global investment would decline by 2 percentage points and its share of global consumption would rise by 5 percentage points.
With consumption growing at roughly 4 percent a year before the pandemic (and much less since), is 6–7 percent growth in consumption possible? No country in history at this stage of the development model has been able to prevent consumption from dropping, let alone cause it to surge, but that doesn’t mean it’s impossible.
But it won’t be easy. With investment growth slowing, which means fewer jobs building bridges, train stations, and apartment complexes, the only way to accelerate consumption growth sustainably is to get household income growth to accelerate through transfers—either directly (such as through wages and other income) or indirectly (such as through a stronger social safety net).
The problem with transfers is that they must be paid for, and there are only three sectors that, in theory, can meaningfully pay for them. One sector that can pay is the rich, who consume a much lower share of their income than ordinary households…
…A second sector that can be forced to pay is the business sector. For example, businesses can pay for these transfers in the form of rising wages, higher taxes, a strengthening currency, or higher borrowing costs (if these are matched by higher deposit rates for household savers). The problem is that with China’s manufacturing competitiveness based primarily on the very low share of income Chinese workers retain relative to their productivity, this would seriously undermine Chinese manufacturing.
The only other sector that can pay is government. There are in fact two levels of government in China: Beijing and local governments. Given the structure of payments and social transfers in China, along with Beijing’s explicit refusal to absorb the various debt and adjustment costs, it is very unlikely that Beijing will be willing to take on the full costs of transfers, which would require mainly central government borrowing.
That leaves local governments as the sector most likely to absorb the costs. By my calculations, if Beijing forced local governments to transfer roughly 1.5 percent of GDP every year to households, it would be possible to drive the growth in both household income and household consumption to around 7 percent annually. This is not as hard as it might at first seem. In spite of terrible cash flow pressures in recent years, local governments may own assets worth as much as 20–30 percent of China’s GDP.
But transferring such a large share of local governments’ assets won’t be easy. Such substantial transfers would be politically contentious and require a transformation of a wide range of elite business, financial, and political institutions at the local and regional level…
…The arithmetic, however, is quite straightforward: unless the rest of the world is willing to reverse its strategic economic priorities to accommodate Chinese growth ambitions, global constraints imply that China cannot continue growing its share of global GDP without sharply reducing the growth rate of investment and manufacturing.
4. The CRISPR Era Is Here – Sarah Zhang
Four years ago, she joined a groundbreaking clinical trial that would change her life. She became the first sickle-cell patient to be treated with the gene-editing technology CRISPR—and one of the first humans to be treated with CRISPR, period. CRISPR at that point had been hugely hyped, but had largely been used only to tinker with cells in a lab. When Gray got her experimental infusion, scientists did not know whether it would cure her disease or go terribly awry inside her. The therapy worked—better than anyone dared to hope. With her gene-edited cells, Gray now lives virtually symptom-free. Twenty-nine of 30 eligible patients in the trial went from multiple pain crises every year to zero in 12 months following treatment.
The results are so astounding that this therapy, from Vertex Pharmaceuticals and CRISPR Therapeutics, became the first CRISPR medicine ever approved, with U.K. regulators giving the green light earlier this month; the FDA appears prepared to follow suit in the next two weeks. No one yet knows the long-term effects of the therapy, but today Gray is healthy enough to work full-time and take care of her four children…
…The approval is a landmark for CRISPR gene editing, which was just an idea in an academic paper a little more than a decade ago—albeit one already expected to cure incurable diseases and change the world. But how, specifically? Not long after publishing her seminal research, Jennifer Doudna, who won the Nobel Prize in Chemistry with Emmanuelle Charpentier for their pioneering CRISPR work, met with a doctor on a trip to Boston. CRISPR could cure sickle-cell disease, he told her. On his computer, he scrolled through DNA sequences of cells from a sickle-cell patient that his lab had already edited with CRISPR. “That, for me, personally, was one of those watershed moments,” Doudna told me. “Okay, this is going to happen.” And now, it has happened. Gray and patients like her are living proof of gene-editing power. Sickle-cell disease is the first disease—and unlikely the last—to be transformed by CRISPR.
All of sickle-cell disease’s debilitating and ultimately deadly effects originate from a single genetic typo. A small misspelling in Gray’s DNA—an A that erroneously became a T—caused the oxygen-binding hemoglobin protein in her blood to clump together. This in turn made her red blood cells rigid, sticky, and characteristically sickle shaped, prone to obstructing blood vessels. Where oxygen cannot reach, tissue begins to die…
…The basic technology is a pair of genetic scissors that makes fairly precise cuts to DNA. CRISPR is not currently capable of fixing the A-to-T typo responsible for sickle cell, but it can be programmed to disable the switch suppressing fetal hemoglobin, turning it back on. Snip snip snip in billions of blood cells, and the result is blood that behaves like typical blood.
Sickle cell was a “very obvious” target for CRISPR from the start, says Haydar Frangoul, a hematologist at the Sarah Cannon Research Institute in Nashville, who treated Gray in the trial. Scientists already knew the genetic edits necessary to reverse the disease. Sickle cell also has the advantage of affecting blood cells, which can be selectively removed from the body and gene-edited in the controlled environment of a lab. Patients, meanwhile, receive chemotherapy to kill the blood-producing cells in their bone marrow before the CRISPR-edited ones are infused back into their body, where they slowly take root and replicate over many months.
It is a long, grueling process, akin to a bone-marrow transplant with one’s own edited cells. A bone-marrow transplant from a donor is the one way doctors can currently cure sickle-cell disease, but it comes with the challenge of finding a matched donor and the risks of an immune complication called graft-versus-host disease. Using CRISPR to edit a patient’s own cells eliminates both obstacles. (A second gene-based therapy, using a more traditional engineered-virus technique to insert a modified adult hemoglobin gene into DNA semi-randomly, is also expected to receive FDA approval for sickle-cell disease soon. It seems to be equally effective at preventing pain crises so far, but development of the CRISPR therapy took much less time.)
In another way, though, sickle-cell disease is an unexpected front-runner in the race to commercialize CRISPR. Despite being one of the most common genetic diseases in the world, it has long been overlooked because of whom it affects: Globally, the overwhelming majority of sickle-cell patients live in sub-Saharan Africa. In the U.S., about 90 percent are of African descent, a group that faces discrimination in health care. When Gray, who is Black, needed powerful painkillers, she would be dismissed as an addict seeking drugs rather than a patient in crisis—a common story among sickle-cell patients…
…Doctors aren’t willing to call it an outright “cure” yet. The long-term durability and safety of gene editing are still unknown, and although the therapy virtually eliminated pain crises, Hsu says that organ damage can accumulate even without acute pain. Does gene editing prevent all that organ damage too? Vertex, the company that makes the therapy, plans to monitor patients for 15 years.
Still, the short-term impact on patients’ lives is profound. “We wouldn’t have dreamed about this even five, 10 years ago,” says Martin Steinberg, a hematologist at Boston University who also sits on the steering committee for Vertex. He thought it might ameliorate the pain crises, but to eliminate them almost entirely? It looks pretty damn close to a cure…
…The field is already looking at techniques that can edit cells right inside the body, a milestone recently achieved in the liver during a CRISPR trial to lower cholesterol. Scientists are also developing versions of CRISPR that are more sophisticated than a pair of genetic scissors—for example, ones that can paste sequences of DNA or edit a single letter at a time. Doctors could one day correct the underlying mutation that causes sickle-cell disease directly…
…We have opened the book on CRISPR gene editing, Frangoul told me, but this is not the final chapter. We may still be writing the very first.
5. Introducing Gemini: our largest and most capable AI model – Sundar Pichai and Demis Hassabis
I believe the transition we are seeing right now with AI will be the most profound in our lifetimes, far bigger than the shift to mobile or to the web before it. AI has the potential to create opportunities — from the everyday to the extraordinary — for people everywhere. It will bring new waves of innovation and economic progress and drive knowledge, learning, creativity and productivity on a scale we haven’t seen before…
…Millions of people are now using generative AI across our products to do things they couldn’t even a year ago, from finding answers to more complex questions to using new tools to collaborate and create. At the same time, developers are using our models and infrastructure to build new generative AI applications, and startups and enterprises around the world are growing with our AI tools…
…We’re approaching this work boldly and responsibly. That means being ambitious in our research and pursuing the capabilities that will bring enormous benefits to people and society, while building in safeguards and working collaboratively with governments and experts to address risks as AI becomes more capable…
…Now, we’re taking the next step on our journey with Gemini, our most capable and general model yet, with state-of-the-art performance across many leading benchmarks. Our first version, Gemini 1.0, is optimized for different sizes: Ultra, Pro and Nano. These are the first models of the Gemini era and the first realization of the vision we had when we formed Google DeepMind earlier this year…
…We’ve been rigorously testing our Gemini models and evaluating their performance on a wide variety of tasks. From natural image, audio and video understanding to mathematical reasoning, Gemini Ultra’s performance exceeds current state-of-the-art results on 30 of the 32 widely-used academic benchmarks used in large language model (LLM) research and development.
With a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU (massive multitask language understanding), which uses a combination of 57 subjects such as math, physics, history, law, medicine and ethics for testing both world knowledge and problem-solving abilities.
Our new benchmark approach to MMLU enables Gemini to use its reasoning capabilities to think more carefully before answering difficult questions, leading to significant improvements over just using its first impression.
Gemini Ultra also achieves a state-of-the-art score of 59.4% on the new MMMU benchmark, which consists of multimodal tasks spanning different domains requiring deliberate reasoning…
…Until now, the standard approach to creating multimodal models involved training separate components for different modalities and then stitching them together to roughly mimic some of this functionality. These models can sometimes be good at performing certain tasks, like describing images, but struggle with more conceptual and complex reasoning…
…Gemini 1.0 was trained to recognize and understand text, images, audio and more at the same time, so it better understands nuanced information and can answer questions relating to complicated topics. This makes it especially good at explaining reasoning in complex subjects like math and physics.
We designed Gemini to be natively multimodal, pre-trained from the start on different modalities. Then we fine-tuned it with additional multimodal data to further refine its effectiveness. This helps Gemini seamlessly understand and reason about all kinds of inputs from the ground up, far better than existing multimodal models — and its capabilities are state of the art in nearly every domain…
…Gemini Ultra excels in several coding benchmarks, including HumanEval, an important industry-standard for evaluating performance on coding tasks, and Natural2Code, our internal held-out dataset, which uses author-generated sources instead of web-based information.
Gemini can also be used as the engine for more advanced coding systems…
…Using a specialized version of Gemini, we created a more advanced code generation system, AlphaCode 2, which excels at solving competitive programming problems that go beyond coding to involve complex math and theoretical computer science.
When evaluated on the same platform as the original AlphaCode, AlphaCode 2 shows massive improvements, solving nearly twice as many problems, and we estimate that it performs better than 85% of competition participants — up from nearly 50% for AlphaCode. When programmers collaborate with AlphaCode 2 by defining certain properties for the code samples to follow, it performs even better…
…We trained Gemini 1.0 at scale on our AI-optimized infrastructure using Google’s in-house designed Tensor Processing Units (TPUs) v4 and v5e. And we designed it to be our most reliable and scalable model to train, and our most efficient to serve.
On TPUs, Gemini runs significantly faster than earlier, smaller and less-capable models. These custom-designed AI accelerators have been at the heart of Google’s AI-powered products that serve billions of users like Search, YouTube, Gmail, Google Maps, Google Play and Android. They’ve also enabled companies around the world to train large-scale AI models cost-efficiently.
Today, we’re announcing the most powerful, efficient and scalable TPU system to date, Cloud TPU v5p, designed for training cutting-edge AI models. This next generation TPU will accelerate Gemini’s development and help developers and enterprise customers train large-scale generative AI models faster, allowing new products and capabilities to reach customers sooner…
…Gemini has the most comprehensive safety evaluations of any Google AI model to date, including for bias and toxicity. We’ve conducted novel research into potential risk areas like cyber-offense, persuasion and autonomy, and have applied Google Research’s best-in-class adversarial testing techniques to help identify critical safety issues in advance of Gemini’s deployment.
To identify blindspots in our internal evaluation approach, we’re working with a diverse group of external experts and partners to stress-test our models across a range of issues.
To diagnose content safety issues during Gemini’s training phases and ensure its output follows our policies, we’re using benchmarks such as Real Toxicity Prompts, a set of 100,000 prompts with varying degrees of toxicity pulled from the web, developed by experts at the Allen Institute for AI. Further details on this work are coming soon.
To limit harm, we built dedicated safety classifiers to identify, label and sort out content involving violence or negative stereotypes, for example. Combined with robust filters, this layered approach is designed to make Gemini safer and more inclusive for everyone. Additionally, we’re continuing to address known challenges for models such as factuality, grounding, attribution and corroboration…
…Starting today, Bard will use a fine-tuned version of Gemini Pro for more advanced reasoning, planning, understanding and more. This is the biggest upgrade to Bard since it launched. It will be available in English in more than 170 countries and territories, and we plan to expand to different modalities and support new languages and locations in the near future.
We’re also bringing Gemini to Pixel. Pixel 8 Pro is the first smartphone engineered to run Gemini Nano, which is powering new features like Summarize in the Recorder app and rolling out in Smart Reply in Gboard, starting with WhatsApp — with more messaging apps coming next year.
In the coming months, Gemini will be available in more of our products and services like Search, Ads, Chrome and Duet AI.
We’re already starting to experiment with Gemini in Search, where it’s making our Search Generative Experience (SGE) faster for users, with a 40% reduction in latency in English in the U.S., alongside improvements in quality.
Disclaimer: None of the information or analysis presented is intended to form the basis for any offer or recommendation. We currently have a vested interest in Alphabet (the company behind Gemini) and Costco. Holdings are subject to change at any time.. Holdings are subject to change at any time.
