저자: ChamathPalihapitiya · 외국인
요약 보기
자율투자(autonomous investing)가 알고리즘 기반 투자의 발전 단계(수동→알고리즘→자동화→자율→에이전틱AI)로 진화해 왔음을 설명. 메달리온 펀드의 연평균 약 39%(수수료 후) 수익을 사례로 제시. 딥러닝·에이전틱AI가 시장을 바꾸고 있으며, 2010년 플래시크래시·2012년 나이트캐피털 사고 등 시스템 리스크와 규제 차이(미국 사후규제 vs 중국 사전승인)를 다룸. 특정 자산에 대한 가격 방향 예측 발언은 없음.
다듬은 글
Today, we're going to unpack one of the most powerful and honestly one of the most secretive forces that's totally reshaping our world. Autonomous investing. And look, this isn't just about trading stocks faster. This is a massive shift in how money, markets, and whole economies actually work. So, in this explainer, we're going to explore this new arms race from the ground up.
You know, where it came from, who the big players are, and the absolutely incredible risks that come with it. Okay, so let's just dive right in with a chart that should well, it should stop you in your trucks. What you're looking at is the annual return of this super mysterious hedge fund called Medallion compared to the S&P 500.
I mean, just look at that performance year in year out. It just crushed the market. We're talking an average of a staggering 39%. And get this, that's after fees. This isn't just someone getting lucky. This is something else entirely. This is the work of a machine. And that right there is the big question we're tackling today.
See, the story of Medallion is really the story of a revolution, a move away from human gut feelings and intuition towards pure cold, hard code. So, how in the world did this happen? How did finance become the playground for mathematicians and physicists instead of the old school stock pickers? For pretty much all of history, if you wanted an edge in investing, it came from who you knew.
maybe some inside information or just brilliant human judgment. But over the last few decades, that edge has completely flipped. Now it's all about the software. We have systems that can sense, decide, and act on way more signals and way faster than any human possibly could. This is a story of how our markets literally became machines.
You know, a really cool way to understand this evolution is to think about it like the levels of self-driving cars. Level one, that's just manual investing. you, a brokerage account, making your own calls. Level two, algorithmic. That's like cruise control, where simple rules execute trades for you.
Level three is automated, handling more complex stuff. But the real leap, where things get wild, is level four, autonomous investing. This is where machine learning models make their own decisions and adapt. And the crazy frontier, level five, is agentic AI. We're talking systems that can plan, reason, and act all on their own, almost like a person, right?
And this journey didn't happen overnight. I mean, it goes all the way back to6002 with the first publicly traded shares. But the real seeds for all this were planted in the 1950s when people started thinking about investing in terms of stats. You know, risk versus reward. Then came electronic trading in the 70s which sped everything up.
Then ETFs in the '90s made it possible to trade an entire index like a single stock. Every single step made the market a little more abstract, setting the stage for the fully autonomous systems we have today. And of course, you can't have this revolution without the intellectual firepower behind it.
It really kicked off with guys like Harry Marowitz and William Sharp in the 50s and60s. They gave us modern portfolio theory, basically a mathematical way to handle risk. But the absolute gamecher came in 1973 with the black skulls formula for pricing options. It took human guesswork completely out of the equation and replaced it with a precise model that really turned options trading into the first corner of finance run entirely by algorithms.
And if there's one person who really embodied this new mathematical approach, it's Edward Thorp. This guy treated investing like an engineering problem. Find a mathematical edge and then control your risk. Period. He actually started out by beating casinos at blackjack and then he just took that exact same quantitative mindset to Wall Street.
His disciplined, model driven style earned him the nickname the godfather of the quants and he totally laid the groundwork for the giants that came after him. Okay, so that's the history. Now, let's jump to today and see how those early ideas have been built into the massive systems that run the show.
We're going to look at the core tech that powers today's markets and the firms that have totally mastered it. This is where we get into how these new titans actually make their money. So, the key thing to get is that these modern systems go way, way beyond just crunching simple numbers. They use tech like deep learning, which is what this neural network diagram is showing.
These models can basically read and understand huge amounts of unstructured data. We're talking news articles, social media posts, you name it. And they pull out these subtle signals that can predict how a price might move. It's basically like teaching a machine to read the mood of the entire market.
And here are some of the firms that are right on the cutting edge. These are proprietary trading shops, places like XTX Markets and Jump Trading. They use these super advanced machine learning methods to trade thousands, even tens of thousands of different things all at the same time. They're not making big risky bets on where the market's headed.
No, they're using deep learning to predict tiny little changes in price. And then they execute millions of trades to capture a really small but really consistent edge. And this is where things start to feel like science fiction. Agentic AI. This is a huge leap. We're not just talking about systems that follow a strategy someone programmed anymore.
We're talking about AI agents that can actually set their own goals, reason their way through complex market data, and then execute these multi-step plans with basically zero human help. They're starting to act like truly independent players in the economy. Now, if you want to see how this is playing out in the real world, check out the Alpha Arena.
It's this live money competition where different AI models are each given 10 grand and just told go trade. And the results, well, they're fascinating. As you can see here, performance is all over the map. One model pulled in a nice 10% return while another one just cratered losing over 12%. It just goes to show you the tech is powerful for sure, but how you build it is absolutely everything.
Okay, so this all sounds amazing, right? But with this incredible power comes some incredible risk. These systems operate at speeds and scales that are just beyond human understanding. So what happens when the code breaks? This is where the story takes a bit of a darker turn as we get into the very real dangers of this revolution.
So the dangers really fall into three big buckets. First, you've got systemic risk. This is when all these different algorithms start interacting in weird ways and create these feedback loops that can throw the whole market into a tail spin. Second is operational accountability. I mean, who's to blame when an AI does something its creators never even imagined?
And finally, societal effects. The risk that this tech just gives a massive, unbeatable advantage to a handful of elite firms, making the gap between the rich and everyone else even wider. Let's look at a real world example of that systemic risk. It's May 6th, 2010, and the market is already on edge.
Then one single large automated sell program kicks into gear. Instantly, highfrequency trading algos see this and pile on, amplifying the selling. Within minutes, all the buyers just vanished. Liquidity completely dried up and the market plunged, wiping out nearly a trillion dollars in value just for a moment.
And the cause? It wasn't one bad actor. It was just the chaotic, unpredictable way all these automated systems bounced off each other. Now for an even scarier story about an operational failure. Back in 2012, a firm called Night Capital, a huge player in the US stock market, rolled out a simple software update.
The error that followed cost them this much, $468 million. And here's the real kicker. They lost all of it in just 45 minutes. The speed of the failure was just mind-blowing. The company, one of the biggest market makers in the country, was pushed right to the edge of total collapse, all before the first hour of trading was even over.
So, how on earth does that happen? Well, it's almost comically simple. A single engineer forgot to deploy a software update to just one of their eight servers. This mistake woke up a piece of old, dormant code that just went completely haywire. A few hundred normal customer orders got sliced and diced into 4 million rogue trades that just flooded the market.
And the most terrifying part of all, they had no kill switch. There was no big red button to stop the bleeding. It was a total catastrophic loss of control. So when you have stories like that, you can bet regulators are going to wake up and pay attention. And that's exactly what happened. It's kicked off this global race to figure out how do we govern these incredibly powerful and complicated machines?
and it's led to two totally different ways of thinking about it around the world. This slide lays out the two competing philosophies perfectly. In the US, the approach is what you'd call expost or after the fact. Basically, firms can deploy whatever they want, but if it breaks the market, they're held responsible.
The focus is on having risk controls and then enforcing the rules. China has gone the complete opposite way with an X anti approach. Regulators there demand to review and actually approve trading algorithms before they're allowed to go live. It's a classic trade-off, right? The US model is all about agility and innovation, while China's is all about safety and control.
So, even with all the risks and all the new rules, this train is not slowing down. The march toward autonomy is absolutely continuing. So, let's look ahead at where this is all going because these systems are expanding way beyond just buying and selling stocks. You can really see the expansion happening across three different layers.
At the bottom, you've got execution autonomy. These are smart systems that figure out the absolute best way to place a trade. In the middle, there's research autonomy. Think of AI analysts like Bloomberg GBT that can read and make sense of all the world's financial news instantly. And then at the very top, you have portfolio management autonomy with these massive platforms like Black Rocks Aladdin that can manage risk across trillions of dollars in assets.
And these are the names you need to know. These are the firms at the absolute heart of this new machine economy. Renaissance Technologies, of course, the ones who started it all, but also giants like Citadel, DE Shaw, Millennium, and Two Sigma. You got to understand, these aren't your grandfather's investment firms.
They're technology companies that just happen to operate in financial markets, and they are pushing the boundaries of what's possible every single day. So, the big takeaway here is this. We've gone from using software as a simple tool to creating systems that are for all intents and purposes independent economic actors.
They can plan, they can act, and they can adapt all on their own. And that leaves us with one final and I think pretty vital question. As these systems get more powerful and more autonomous, who or what is truly in control?
Comments 0