How does Bridgewater invest in AI?

Author: Cao Zexi

Image source: Generated by Unbounded AI

What does the world’s largest hedge fund think about AI?

On Monday, July 3, Bridgewater contacted Chief Investment Officer Greg Jensen in an interview to systematically talk about Bridgewater’s views on AI technology, and shared his views on how Bridgewater invests in AI, how to use AI investment, and its outlook on AI technology, etc. perspective on the problem.

How Bridgewater invests in AI

Jensen name:

In the process of reorganizing Bridgewater, we also did something that we hadn’t done before, which is to let some people research and invest in things that may not be profitable immediately, but this is our long-term project. Therefore, we set up this AI project with a team of 17 people led by me. I am still actively involved in the core work of Bridgewater, but the other 16 people are 100% committed to reshaping Bridgewater through machine learning. We’re going to have a fund that’s run exclusively on machine learning technology, and that’s what we’re doing in the lab right now, pushing the limits of AI, machine learning capabilities. Right now, it is a big problem to set up such a fund. If we take large language models, they have two types of problems. One, these models are trained more on the structure of language, so they often return something that looks like it’s structured, grammatically correct, but not always an accurate answer. this is a problem. Second, it hallucinates, it makes things up because it pays more attention to the structure of the word or concept that comes next than to the accuracy of the concept that comes next.

Therefore, Jensen believes that AI can help people conceptualize and theorize what they observe, but there is still a long way to go before using AI to choose stocks. So Bridgewater’s real focus is:

**But there are other ways to combine this with statistical models and other types of AI. That’s what we’re really focusing on, combining large language models that are less accurate with statistical models that are good at describing the past accurately but are terrible at predicting the future. ** **Bringing these together, we begin to build an ecosystem that I believe enables what the Bridgewater analysts are doing. ** If this ecosystem is completed, we will have millions of investment partners at the same time. If we had the ability to control the hallucinations and errors of AI through statistics, we could do a lot of work quickly. That’s exactly what we’ve done in the lab and demonstrated that the process works.

How does Bridgewater invest through AI?

If you can build an ecosystem that includes AI and other technologies, how will Bridgewater use this system to invest?

Jensen believes that statistical AI and large-scale language models can complement each other and play the role of Bridgewater’s “left and right hand” in investment:

Statistical AI can take theories, go back to whether those theories were true at least in the past and what their flaws were, and refine them, offer advice on how to do things differently, and then we can have a dialogue with them. One advantage that large-scale language models have is taking a complex statistical model and talking about what it’s doing. There are ways to train language models to do this. The way we model this is that language models can come up with underlying theories. It’s not the most creative thing in the world, but it’s theory at scale, that’s for sure. Again, large-scale language models are great, but we have to tune the language model in some way, and we can use statistics to control it. We can then use the language model again to take the results in the statistical engine and discuss it with a human or other AI and report what was found, what and what kind of theory. If the conclusions reached are contrary to people’s perception, then conduct more tests. This is the cycle I’m very excited about, as I said, statistical AI has been limited so far because it’s focused on market data. For language models, the benefit is that it can better understand things that statistical models do not. For example, statistical models of markets have no concept of greed, but large-scale language models can almost understand the concept of greed - these models have read everything about greed and fear and such. So combining the two now produces a human-like mode of thinking.

What does AI mean for human employees?

Over time, computers can do more and more things. Jensen believes:

What I want to say is that today, humans have become accustomed to only fulfilling roles related to intuition and creativity. We use computers to memorize and run these rules continuously and accurately. It’s only halfway through the transition, and now there’s another leap forward. There is no doubt that AI will change the role of investment assistants. Exactly, we still need people to work around these things for the foreseeable future, we still need a while to build out the ecosystem of these machine learning agents and so on. **Leveraging AI will be part of the future of work and I think it will be difficult in any knowledge industry not to leverage these technologies. ** In terms of computer programming, we are seeing a huge breakthrough in coding. Now, with AI, people only need to know what to code, not how to code, which is a huge breakthrough. So a bunch of people who don’t have great training or ability in C++, Python, or whatever can suddenly get what they want much faster. **So all of a sudden, the skill sets needed in the workplace are changing, and the way they’re changing is surprising to many people because it’s actually a lot of knowledge work, like content creation and so on, and people at one point It is thought that the time to be replaced by machines is still in the distant future, but in fact it is just around the corner. ** So the bottom line is that there is so much change now that it is imperative to have flexibility in the workplace and be able to leverage any tool.

Can AI be used to directly manage investments?

There are a variety of AI investment management tools on the market. What people are concerned about is, with the great development of AI, will human beings only need to invest in AI in the future?

Jensen believes:

I thought it would both cause an accident and get me really excited. Obviously, I’m excited about the power of AI, and I think there are ways to put it to good use. But at the same time, AI makes a lot of mistakes. Some funds use GPT to pick stocks, but these fund managers do not really have a deep understanding of AI and possible weaknesses. In one example, in the real estate market, Zillow, a real estate brokerage platform, uses AI technology to predict house prices, evaluate house prices, and enter the market to start buying houses that AI believes are undervalued. However, Zillow has several problems. One is that while they have a lot of housing data, it happens over a relatively short period of time. So while they have a seemingly large number of data points, there is still a macro cycle that influences the assessments they make. Second, they underestimate the disconnect between theory and practice when it is actually an adversarial market. So obviously this is a huge problem for Zillow, they had a big impact on the real estate market and then a huge failure. Going back to the stock market, very short-term trading, arguably more suitable for machine learning, because there is a lot of data, and AI can learn faster through this data. But on the other hand, in the longer term, the role of AI may not be able to play out. Data is often like heart rate data for a person’s lifetime. You might be thinking, wow, my heart has been beating for 49 years, that seems like a lot of stats, but when you have a heart attack, it’s completely irrelevant. So, even with large amounts of data, it can be misleading, and these problems cause huge problems for these techniques. **So one has to understand the tools, what they are good at and what they are not good at, and combine them in a way that takes advantage of the strengths of each tool and avoids the weaknesses. ** There is still a lot of work to be done on large language models, which we can certainly train with reinforcement learning to make sure they don’t make known mistakes.

Are markets still dominated by optimism?

Jensen believes that the market is still dominated by optimism. He said:

The Fed appears to be a little more realistic than the market on what it will do. When you look at the market’s reaction, it’s very bullish. But we have to note that, historically, the market tends to be too optimistic.