Marks states nobody can specify what AI will do, when, for whom, or how much profit it will produce.
“I've never heard anybody tell me exactly what AI will be able to do or when or for whom or how much profit it'll produce and for whom.”
Marks wrote that if AI exuberance doesn't produce a money-losing bubble, it will be the first technological innovation not to.
“So I wrote in a memo recently this year, and I think it's true that if this technological innovation with its exuberance doesn't produce a money losing bubble, it'll be the first.”
Marks calls AI the hardest thing he has ever seen in investing due to enormous uncertainty.
“This is the hardest thing I think I've ever seen in the investment world because of this enormous degree of uncertainty.”
Marks describes Claude making a wordplay joke when asked to be hypercritical of his memo.
“So it writes me back, and it says, do you want me to be hypercritical or hypocritical? It's making a joke.”
Marks reports Claude made a joke distinguishing between hypercritical and hypocritical.
“So it writes me back, and it says, do you want me to be hypercritical or hypocritical?”
Marks argues AI innovation speed with Claude and coding models is faster than any previous technology.
“I think innovation, I think if you look at the Claude and all the coding models and the way they've progressed and the way and the way Anthropix revenues have progressed, I think you have to say that this is faster than anything we've ever seen before.”
Marks warns AI's innovation speed exceeds society's adjustment capacity, predicting significant dislocation period.
“Mhmm. So you might say it'll catch up, but I think you could I think at minimum, you're talking about a significant period of dislocation.”
Marks argues AI differs from prior technologies by designing jobs autonomously rather than just doing assigned work.
“AI, it's it's different. It's not just gonna do the job we used to do. It's gonna design new jobs. Mhmm. It's gonna assign new jobs.”
Marks argues AI differs from prior technologies by autonomously designing and assigning new work without instruction.
“It's gonna take on work we didn't think it could do, and it and and it's gonna operate, at some point in time without instruction.”
Marks notes ChatGPT's new model helped design itself, which he finds unprecedented.
“And in the write up for the for the model, they said, basically, in English, AI, the model helped us design the model.”
Marks questions whether AI can identify future Amazon or Steve Jobs from present information.
“Can AI sit down with five business plans and figure out which one is Amazon? Can AI sit down with five CEOs and figure out which one is Steve Jobs.”
Marks questions whether AI models trained on same history can produce different investment outcomes.
“Are some AI models smarter than others? Since they all have super IQs, computing power, and they all are trained on the same history, are some smarter than others?”
Marks argues government can replace paychecks but not the sense of purpose and structure that work provides.
“And and the government, in theory, can make up the paycheck, but they can't make up the sense of purpose and the reason to get out of bed and the structure for your day.”
Marks doubts AI can pick the best managers because it requires intuition and subjective feel.
“Because picking the best managers requires an intuition and a subjective feel that I would be surprised to learn that AI can be very good at.”
Marks argues AI makes the world more unpredictable than any time in his lifetime, challenging investment decision-making.
“the changes that are underway today, and in particular the introduction of AI, render the world much less predictable than at any time, probably any time ever, and certainly any time in my lifetime.”
Marks cites Block eliminating 40% of its workforce in one day as evidence most people underestimate AI's impact.
“40% of the workforce gone in one day because AI could do the work cheaper and faster. So how many people in the world understand the potential import of that?”
Marks notes OpenAI revealed their new model helped build itself, which he finds eye opening.
“OpenAI brought out a new model. And in the descriptive materials, they said that the that the model helped build the model, which is really eye opening.”
Marks says AI can eliminate a huge percentage of knowledge work, citing Block laying off 4,000 people.
“Clearly, it can eliminate a a huge percentage of of knowledge work. And you we saw Friday block 10,000 employees, let 40%, 4,000 people let go”
Marks says AI can eliminate a huge percentage of knowledge work, citing Block laying off 4,000 employees in one day.
“And you we saw Friday block 10,000 employees, let 40%, 4,000 people let go in one go, who who aren't needed anymore because AI can do it better and cheaper and faster.”
Marks expresses concern that AI moves faster than society can adjust, creating a formula for disruption.
“So one of my concerns is that AI moves faster than the ability of society to adjust to it. And that and that is a formula for disruption.”
Marks warns AI may move faster than society can adjust, creating a formula for disruption.
“one of my concerns is that AI moves faster than the ability of society to adjust to it. And that and that is a formula for disruption.”
Marks says if you think you know what will happen with AI, you don't understand what's going on.
“I think it was Walter Cronkite who said if you're not confused, you don't know what's going on. I would say if you think you know what's gonna happen,”
Marks argues anyone claiming to know AI's trajectory doesn't understand the situation; we're in the first inning.
“I would say if you think you know what's gonna happen, you don't know understand what's going on. We're at the we're in the first inning of a very long unpredictable game.”
Marks says if you think you know what will happen with AI, you don't understand what's going on.
“I would say if you think you know what's gonna happen, you don't know understand what's going on.”
Marks describes AI as the first inning of a long unpredictable game with unknown rules.
“We're at the we're in the first inning of a very long unpredictable game. We don't know what the rules are or or have any idea how many innings there are”
Marks says we're in the first inning of a very long unpredictable AI game with unknown rules.
“We're at the we're in the first inning of a very long unpredictable game. We don't know what the rules are or or have any idea how many innings there are in the game.”
Marks cites Thinking Machine Labs raising two billion dollars at twelve billion valuation without disclosing its product as bubble behavior.
“some woman left OpenAI, started a company called Thinking Machine Labs, went out to raise money, and she said this company is going to engage in AI, but I can't tell you what we're going to do. It's a secret. And people gave her $2,000,000,000 for a sixth of the company.”
Marks says nobody can explain how AI will change the world, unlike the internet bubble where the vision was clearer.
“I've never heard anybody tell me how AI is going to change the world. We know it's a powerful force. Can think, it can process data.”
Marks cites Buffett's point that productivity gains from internet and AI may not translate to profitability, now applied to AI.
“There's no doubt that the internet will produce a great increase in productivity. It's not clear that it'll have a positive impact on profitability. And I think the same is true of AI.”
Marks quotes Buffett distinguishing productivity gains from profitability in technology like internet and AI.
“There's no doubt that the internet will produce a great increase in productivity. It's not clear that it'll have a positive impact on profitability.”
Marks notes AI could eliminate half of entry level jobs but questions whether productivity gains translate to profitability.
“You say that AI has the ability to eliminate half of entry level jobs. That was the whole conversation because then they cut to something else. But the point is that may be true.”
Marks notes CNN reported AI could eliminate half of entry level jobs while maintaining US GDP, raising profitability questions.
“And obviously if you can produce The US GDP and eliminate half the entry level jobs, it could be more profitable or certainly more productive.”
Marks questions whether AI savings will accrue as profits or be competed away through lower prices.
“To whom will the savings accrue? If different companies are competing to provide the AI service, maybe they'll compete on price to the point where it's not profitable for them.”
Marks suggests AI savings may accrue to consumers through price competition rather than to company profits.
“If different companies are competing to provide the AI service, maybe they'll compete on price to the point where it's not profitable for them.”
Marks frames the AI investment choice as binary moonshot bets versus incremental gains in established tech companies.
“Or do you want to invest in a great tech company, which is already existing and making a lot of money where AI could be incremental, but not life changing?”
Marks cites Buffett's distinction that productivity gains from Internet and AI may not translate to profitability.
“There's no doubt that the Internet will produce a great increase in productivity. It's not clear that it'll have a positive impact on profitability. And I think the same is true of AI.”
Marks questions whether AI eliminating half of entry-level jobs will translate into profits or just lower consumer prices.
“if you can produce The US GDP and eliminate half the entry level jobs, it could be more profitable or certainly more productive. But the question is, will it be more profitable?”
Marks is certain AI will irreversibly change society but questions whether its implementation will prove excessive in scope and financing.
“And the question is, will the implementation prove to have been excessive in scope and in in the way it's financed?”
Marks argues job losses from automation and offshoring coincided with the opiate epidemic in both amount and geographic location.
“And as I said in the addendum, When we lost jobs to automation and offshoring, I think that that coincided with the opiate epidemic and not only in amount but also in location.”
Marks says current AI frenzy has not reached bubble-level mania yet.
“And I don't detect that level of mania at this time, so I have not put the bubble label Right.”
Marks says he has not labeled current AI frenzy a bubble because mania has not reached critical level.
“And I don't detect that level of mania at this time, so I have not put the bubble label Right. On this on this incident.”
Marks says AI will change the world and has been successful, with investors piling in.
“I think there's relatively little doubt that AI will change the world. And AI has been successful as an investment, and people are piling in,”
Marks says AI will change the world and has been successful, with investors piling in amid FOMO.
“I think there's relatively little doubt that AI will change the world. And AI has been successful as an investment, and people are piling in, and there's some fear about being left out.”
Marks judges that AI investing has not reached the critical mass of mania needed for a bubble.
“But to me, it just hasn't this is a judgment call. And to me, it just hasn't reached that critical mass of mania.”
Marks argues computers cannot identify the next Steve Jobs or Amazon through subjective future judgments.
“I don't think it can look at five business plans and figure out which is the next Amazon. These are subjective judgments about the future,”