Why AI Can’t Beat the Market
I thought this was funny: “We Asked ChatGPT to Make a Market-Beating ETF. Here’s What Happened.”
One of the reasons I got out of professional public markets investing was the fact that lots of technology – including artificial intelligence – was being thrown at the stock market in the neverending quest for outperformance. Interestingly, though, I don’t think such advances make it likely for someone to achieve better returns for the same reason that ChatGPT wasn’t able to design a market-beating portfolio. Rather, I think such things make the market much, much more efficient and therefore decrease the probability that anyone anywhere is likely to earn an outlier return.
That’s because there is no optimal investment strategy. Instead an investment strategy that works works only in the context of all of the other investment strategies going on around it. If everyone has trained algorithms to look for great companies based on certain quality factors, then those stocks will necessarily be expensive and that will definitionally be a bad investment strategy. If, on the other hand, no one wants to buy small private businesses, then those businesses will likely be cheap and that might be a strategy that earns good returns. The rub is that those things won’t be true forever. When people see others earning good returns on small private businesses, they will look there and that will become a bad investment strategy and when they stop buying great public companies, that will become a good investment strategy.
That’s why you want to be idiosyncratic; to zig when others zag.
AI, I think, is trained on data, so it does fantastically in areas when truths are immutable. Investing is anything but. The only things that consistently work when it comes to capital allocation are pragmatism and patience.
– By Tim Hanson