AI Will Make You a Better Investor, Not a Great One
AI can take a weak investor to competence. The jump from competence to greatness is the part it can't help with.
In 2023, two MIT researchers assigned a writing task to 453 college-educated professionals. Half of them used ChatGPT. The AI group finished 40% faster and scored 18% higher. Everyone expected that AI would increase their productivity, and it did.
The gains weren’t even. The weakest writers improved the most, while the strongest didn’t write any better with AI at all. They simply finished faster. The tool pulled the bottom toward the middle and left the top where it was.
I’ve used AI to research and manage my own portfolio for the past year. It’s fast at the busywork and useless at the only thing that pays: catching what the market got wrong. Now, anyone can run that type of research. Being skilled isn’t enough anymore. The only investors worth hiring are those who can outsmart the machine.
Everyone gets the same answer now
I called AI a consensus machine in AI for Investment Research. It tells you what people think about a company. It does this based on the training data it learned from. A couple of years ago, getting that kind of read was tough. You had to read the filings, build a model, and understand the business to form an opinion. Now you can hand a 10-K to a chatbot and have most of it done in twenty minutes. If you had never done it before, that’s a real leap. You go from knowing nothing to a solid, middle-of-the-road grasp of a company over lunch.
The catch is that everyone else’s chatbot produces the same solid, middle-of-the-road grasp. And it’s worse than common because the answers converge. Doshi and Hauser, researchers in AI and creativity, asked people to write short stories. Some got help from a model, while others wrote on their own. The AI stories did better. This occurred as a result of the improvement in the weaker writers. The strongest writers saw no change at all. But the AI stories also started to resemble one another. Each writer did better work, and the work blurred together.
That convergence is what matters for investing. Now, anyone can write a decent thesis on a stock. But they all end up with similar ones. This happens because they use the same model, which is based on the same consensus.
How to tell if your edge is real
Back when a good thesis was rare, having one was worth a lot. A decent thesis is now free, and they all sound the same. So, decent work has little value. People will only pay for a unique, correct viewpoint.
Check the stocks you chose. Go through the list and ask if a chatbot would have suggested buying each one. Anywhere it would have said yes, you’re holding the consensus, and the consensus is free now. The picks that matter are the ones a chatbot would have flagged. It’s the stock nobody wanted. You held it for a year and a half, convinced the crowd was wrong.
If the honest answer is “a chatbot would have said buy” every time, move that money to index funds. You’re not giving anything up because the analysis you were doing is free to everyone now anyway.
What to read next
📖 Range by David Epstein. The argument is that people with broad, mixed backgrounds often catch what the specialists miss. Useful for thinking about what’s left for humans once the narrow technical work gets automated.
📖 The Almanack of Naval Ravikant by Eric Jorgenson. Naval separates the knowledge you can’t easily teach or copy from the kind anyone can pick up. Same split as the one between a real edge and a chatbot answer.
📖 The Outsiders by William Thorndike. Eight CEOs who beat the market by ignoring the conventional wisdom of their day. Worth reading for how often the right move looked wrong at the time.
🎧 All three are excellent on Audible. The free trial gives you one credit to start.
As an Amazon Associate, I earn from qualifying purchases.