Introduction
A lot of questions have been swirling lately about the impact of artificial intelligence and machine learning on the art (and business) of investing. Some people have argued that independent investment managers like Aquamarine will go the way of the dodo — a casualty of AI, increased computing power, and machine learning. I want to discuss this briefly because I think it might be useful in terms of helping you to understand my idiosyncratic business strategy and how you fit into it.
First, let’s talk about AI. I want to share with you a little bit about my (meager) understanding of some important concepts related to computer science — albeit, with the caveat that I am neither a computer scientist nor a mathematician.
We live in a world where dramatically increased computing power is being applied to solve a panoply of problems, including image recognition, driverless cars, and stockpicking. Does that mean that we’re moving into a world where there will be no space for stockpickers and investment managers? I don’t think so.
What I have learned in my forays into this subject is that computer problems can be categorized according to different classes of complexity. And it turns out that it’s not hard to define a simple problem and to prove mathematically that, in practical terms, it would require an almost infinite amount of computing power to solve it. Of course, many problems are becoming solvable with increases in computing power, and these advances can make our lives immeasurably better. But there are some problems that are likely never to be solved by computers.
With this in mind, let’s think about the task of creating a form of AI that could replace the independent stockpicker. In order to learn, the AI has to train itself on a data set. In the case of Alpha Go, the rules of the game were simple enough that the AI could play itself in order to create its own data set. But that is a game whose rules are known.
The world of human affairs is infinitely more complex. Humans, corporations, raw materials, economic conditions, markets, and prices are all interacting. We don’t know the rules. So we need to learn them by using a data set. But when it comes to stockpicking, the data required might include all economic history, all annual reports, all regulatory filings, all news releases, and so on. In other words, the magnitude of the stockpicking problem is almost unimaginably immense. By comparison, developing driverless cars is a cinch because the rules are relatively easy to define. Yet that’s a problem that is still nowhere near to being solved.
Now, I don’t doubt that sophisticated computer programs can be designed to take advantage of minor price movements that occur over milliseconds and to shave off minuscule gains based on those micro-movements. But the challenge of designing computer programs to replicate the long-term success of, say, Buffett, Munger, Seth Klarman, or Li Lu is quite another matter.
People in Silicon Valley are intoxicated with the idea that computing power can solve every problem on the planet. But I suspect that the investment challenge will never be solved to the satisfaction of these techno optimists, partly because they underestimate the immense complexity of the problem and the quantity of data required to solve it. The best answer to this problem is not machine intelligence, but human intelligence.
Just ask yourself for a moment: What makes Buffett so successful? How has his system at Berkshire generated so much wealth? It’s not just that he and Munger are intellectually brilliant or have worked hard to deserve their success. It’s also, in part, that they have generated enormous amounts of goodwill over a lifetime of investing. Thanks to the power of reciprocation, they have a global set of brains working for them. There are countless people (me included) who are so grateful that we wouldn’t hesitate to help them if we could. And this goodwill system has been amplified by investing in businesses that are operating in the same manner — delivering enormous value to others and acting with admirable integrity.
It’s this human factor that lands Buffett with the best ideas, the best people, the best deals. What better moat could there be? These all-important human traits and behaviors are impossible to replicate with any computer program — and yet they are absolutely central to Buffett’s and Berkshire’s success.
But why is any of this relevant to you? It’s relevant because what we are attempting to do at Aquamarine is to build exactly the same type of moat. We want to be successful because we deserve to be successful, because we deliver enormous value to you, and because we operate with total integrity. When I think about how to build our competitive moat, it’s not just about buying high-quality businesses. It’s about how we behave. It all starts with the most basic moral principles. For example:
- Tell the truth.
- Mean what you say and say what you mean.
- Treat people well.
- Keep what is told to you confidential.
- Don’t tell people what to do but share the benefits of your own experience.
- Seek to deliver value to your stakeholder groups before extracting value for yourself.
- When extracting value, extract only a small proportion of the value that you have created for others.
The people I admire most in the world of business and investing embody this kind of behaviour. That includes Buffett, Munger, Peter Kaufman, and others. Not surprisingly, they tend to be found around companies that also embody this way of operating, such as Berkshire Hathaway, Costco, and others.
But our goal at Aquamarine is to protect and grow the money that has been entrusted to us. Telling the truth and trying to be decent people is not enough to protect you from the many ways in which the world wants to separate you from your wealth. For example, there are trading costs, custody costs, taxes, the economic environment, behavioural uncertainties, and the difficulty of selecting the right investments. All of this has led me to think more broadly about how to create value for you.