저자: KenGriffin · 외국인
요약 보기
AI가 6~8주 걸리던 PhD급 금융 논문 재현·검증 작업을 약 2~3시간에 끝냄. 기업들의 AI 성공 서사는 과장된 경우가 많음. AI는 고숙련 전문직을 압박하지만 창업 기회를 확대하고, 생산성 향상이 곧바로 감원으로 이어지지는 않음. 단기 정량 우위는 대체 데이터로 빠르게 소멸해 장기 통찰력의 가치가 커짐. 컴퓨트 수요 급증으로 연산 자원 가격이 구조적으로 상승 중이며 이는 석유처럼 중요한 자원이 됨. 기술 속도와 제도적 적응 속도의 격차가 불안의 원인.
다듬은 글
Hey, a task that used to take a team of PhDs 6 to 8 weeks just got finished by a machine in under 3 hours. And the man who watched it happen says that was the moment he went home and felt genuinely shocked and depressed. Stay with me because what caused that reaction isn't what you'd expect. And by the end of this, you'll understand exactly why it should worry you, too.
A couple of years ago, a dinner brought together leaders from some of the biggest global companies in the world. Every single person at that table was buzzing with excitement about how AI was transforming their business. So, a simple question got asked. Can you actually explain how specifically AI is driving these results?
And here's the twist nobody saw coming. . Every single one was , technology that had existed for years, just repackaged with a trendier name. In boardrooms across the country, the line between AI and technology had gotten dangerously blurry. And a lot of the hype in corporate America was built on that confusion.
Here's the first lesson. Before you believe any company's AI success story, ask for the specifics. More often than not, the real answer is far less dramatic than the headline. But there was one story from around that same time that didn't fit that pattern at all. And that one is the story that actually mattered.
Here's the moment that changed everything. A team built an AI system with one specific job. Take an academic finance paper, reproduce the underlying research from scratch, and verify whether the results actually held up. This is dense technical work. Academia publishes an enormous volume of papers in finance, and reading through them, testing the hypothesis, judging the quality of the methodology, asking whether the findings would hold up outside the original sample, that work normally somewhere between six and eight weeks per paper.
It's genuinely interesting work and even just a handful of solid ideas pulled from these papers can be worth real money. The AI system did the entire process in roughly two to three hours. Sit with that for a moment because this wasn't some simple task getting automated. This wae kind of job that represents years of specialized training compressed down into an afternoon.
Here's the second lesson. The real shock with AI isn't that it can write an email or summarize a document. It's that it's now capable of doing deep specialized intellectual work that used to define entire careers at a fraction of the time. And here's the detail that makes this story even more unsettling.
This breakthrough didn't come with any job cuts at all. Let's talk about why that's actually the complicated part. Here's something that might surprise you. Despite this breakthrough, there was no reduction in head count. None. Why? Because there's simply an enormous number of problems still worth attacking.
Every productivity gain gets absorbed, not by cutting people, but by giving the same talented team more ground to cover. That sounds like good news. And in one very real sense, it is. But flip it around and there's a second part of the story that's genuinely alarming for an entire industry. Competitive advantages that used to take years to build are being filled in at breathtaking speed.
Think about what that means for very high-skilled work across the economy. In certain areas, that transition is especially difficult. Translation work is a clear example. If your profession is translating one language into another, that skill is facing direct immediate pressure and retraining for something else entirely isn't simple or fast.
So, you end up holding two completely different thoughts in your head at the exact same time. On one hand, the impact of this technology on very high-level skilled work in the job market is genuinely alarming, especially for people whose entire careers were built around a specific specialized skill.
On the other hand, this is also quietly creating a golden age of entrepreneurial opportunity. Here's the third lesson. A technology that doesn't cost anyone their job today can still be quietly eliminating the career path they were counting on for the next 10 years. And that second part, the opportunity side, might be the single most important shift happening in the entire economy right now.
Here's what that actually looks like. Here's the part of this story that deserves far more attention than it usually gets a large team. Entrepreneurs are about to be able to launch new businesses at a speed that simply wasn't possible 5, 10, or 20 years ago. They're going to be able to take on established incumbents in ways that would have seemed unthinkable a decade ago.
Here's a real example. A friend of a friend built a startup that, under normal circumstances, would have needed 30 or 40 employees just to function. They couldn't afford that many people. So, how do they actually run the business with just a handful? Agentic AI systems doing the work that used to require an entire department.
Over the next couple of years, expect to see a lot more stories exactly like this one as entrepreneurs embrace this technology to meet customer needs in ways that create real value without needing the massive headcount that used to be a prerequisite for competing at scale. Here's the fourth lesson. The same force that's quietly erasing certain specialized careers is simultaneously handing a single sharp-minded founder the leverage that used to belong only to large, well-funded companies.
So, if this is reshaping entire industries this fast, what does it mean for something as fundamental as how people actually get hired and paid to think for a living? That's where this gets genuinely complicated. Let's talk about a very specific kind of work, picking stocks, evaluating companies, making long-term judgment calls about what's actually valuable.
That kind of work has always been considered a timeless business, a skill set that doesn't really go out of style. The question worth asking is, what does that same work look like in the future, now that AI can compress 6 weeks of PhD-level research into 3 hours? Here's the honest answer. The core skill set stays remarkably similar.
But the market will increasingly reward a very specific kind of insight, genuinely understanding which companies are creating transformative products that will actually change society far more intensely than it rewards simply predicting whether a company beats this quarter's earnings estimate. And here's why that second part has gotten so much harder to use as an edge.
The rise of alternative data has made the here and now of a company's performance far more transparent and far more quickly understood by everyone at once. Access to things like aggregated consumer spending data can tell you what a company's quarterly revenue is going to look like well before the actual earnings report comes out.
That's been a decade-long transformation, but it's accelerating now because of the combination of talented people and increasingly capable AI technology working together. Here's the fifth lesson. When the short-term, easily quantifiable edges get erased by widely available technology, the only real competitive advantage left is long-term vision, the ability to see years ahead of where the data currently points.
All of this, the compute, the automation, the shifting value of human judgment, runs on a single resource that most people never think about. And that resource is where this story gets genuinely tense. Every piece of this AI story runs on one thing, compute. The raw processing power sitting inside massive data centers, running around the clock, machine learning has been a critical part of this industry's story for a full decade now.
The rise of transformer models over the last several years has continued to improve the ability to both price risk and manage it accurately, not just for one firm, but across the leading market making operations in the entire industry. Here's the fact that reframes everything. Essentially, all of the available compute capacity in the world right now is already being used all the time.
There's no idle supply sitting around, which means the real question isn't who has access, it's who's willing to pay the absolute most for it. And the price per unit of compute has climbed well beyond where anyone would have reasonably projected it just two or three years ago. Firms that rely heavily on these tools are now spending hundreds of millions of dollars on compute.
That's simply the going market rate for this capability today. Just like the price of jet fuel climbs during a shortage, there's real structural inflation happening in the cost of computing power itself. And firms with thinner margins simply can't absorb that cost the way firms with stronger margins can.
Here is the sixth lesson. In this new economy, compute is becoming as foundational a resource as oil was in the last century. And whoever controls that supply holds serious leverage over everyone who depends on it. Now, let's bring this back to the emotional core of this whole story, because that's really what started this entire conversation in the first place.
Here's the honest answer to the question everyone eventually asks. Why did a single Friday evening leave someone feeling shocked and depressed about AI instead of excited? It wasn't about the technology failing. It was about the technology succeeding too fast, faster than the systems built to support the humans affected by it.
Six to eight weeks of specialized PhD level work compressed into an afternoon isn't just an efficiency story. It's a preview of how quickly entire categories of human expertise can be compressed while the institutions responsible for retraining people, universities, labor policy, corporate HR departments are still operating at the old much slower pace.
That mismatch between how fast the technology moves and how slowly humans and institutions can actually adapt is the real source of the unease, not science fiction fears about machines taking over. A much quieter, more immediate fear that the gap between those who adapt in time and those who don't is widening faster than anyone in a position of power is prepared to address.
Here's the seventh and final lesson. The right response to AI isn't blind panic and it isn't blind excitement either. It's clear-eyed urgency recognizing that the window to build new skills and reposition yourself is shrinking faster than most people currently realize. Let's bring this all together.
It started with a dinner where nobody could actually explain what AI was doing for their business. It led to a six-week research task completed in three hours, a golden age of entrepreneurship hiding inside the disruption, a shifting definition of what human judgment is even worth anymore, and a compute market now more valuable and more contested than oil once was.
None of this is a reason to panic. But it is exactly the reason behind that one honest, uncomfortable feeling, shocked and depressed, not because AI is evil, but because it's moving faster than any of us are truly ready for. If this helped you understand the real story behind the AI headlines, , does AI speed excite you more or worry you more? See you in the next one.
Comments 0