A.I.'s Investment Illusion: The Mirage of Market Mastery

A University of Chicago economist told us as much 102 years ago.

The evidence is in.

AI-powered funds fail to match the performance of human portfolio managers.

I can't say I'm surprised.

A.I.-powered funds were sold as the Holy Grail investing.

But like all other efforts to beat the market by churning and chopping ever finer data points, it has fallen short.

University of Chicago economist Frank H. Knight told us as much 102 years ago.

It all comes down to understanding the distinction between two crucial concepts: “risk” and “uncertainty.”

Yet, it’s a distinction I don’t recall ever coming across in standard financial texts.

But once you understand it, you’ll never look at claims of “market-beating strategies” again.

Let me explain…

Frank Knight, University of Chicago

A.I. Investing’s Missteps in 2022

Conventional wisdom will tell you three factors that explain A.I.'s recent underperformance.

First, AI lacks the nuanced judgment of experienced investors.

A.I.'s algorithm rigidly avoided Meta Platforms (META)-ironically, a leading A.I. stock. It was just too expensive.

As a result, A.I. funds missed out on huge gains.

By definition, A.I. looks back, not forward. A.I. cannot account for qualitative factors that alter a stock's prospects.

This glaring blind spot hampers returns.

Second, A.I.'s memory is far too short to handle market shocks.

2022's historic interest rate hikes exposed this weakness in A.I. funds. A.I. cannot adapt to sharp shifts in volatility without sufficient data encompassing diverse market conditions. A.I. just needs to get smarter.

Finally, A.I. makes investment mistakes no human would.

A.I. fund’s returns suffered from poor position sizing. Funds spread bets too thin instead of concentrating on market trouncing tech leaders.

An experienced manager would know to take a more concentrated approach in strong sectors. Alas, A.I.'s errors prove it cannot (yet) match human intuition.

In sum...

Deficient judgment, inadequate training, and non-human errors explain why A.I. funds fall short of top asset managers.

But that will change- or so we are told.

After all, A.I. is in its infancy. Just give A.I. time to learn from experience.

I call B.S.

Here’s the reality.

Financial markets are far too complicated to be shoehorned into mathematical equations- AI-driven or not.

And financial markets are far more difficult than chess- or even Go- two of humanity’s most complex games that A.I. has conquered.

Risk vs. Uncertainty: A.I.'s Unwinnable Chess Match with the Market

AI’s attempt to solve complex problems, like predicting the market, is not something new.

Nor is conflating two distinct but related concepts like risk and uncertainty.

In 1921, Frank Knight of the University of Chicago wrote an entire book on the difference between risk and uncertainty in Risk, Uncertainty, and Profit.

“Risk” can be measured and quantified through theoretical models. That allows you to predict a range of outcomes. (Think chess)

In contrast, “uncertainty” cannot be expressed in quantitative terms. Future events are unpredictable. (Think financial markets)

Here are three more distinctions to hone our understanding.

First, all the potential outcomes are known in risk. Chess ends with a checkmate.

The number of ways to get there is vast. But it is finite.

In case of uncertainty, the outcomes are unknown. No amount of computing power can calculate a future outcome with any certainty.

The magnetic poles of the earth may invert tomorrow, destroying human civilization. But I bet not even Renaissance Technologies has that contingency embedded in their models.

Second, you can manage risk through proper measures. Bet small. Play defense. Implement a better strategy. 

But outcomes in the world of uncertainty are beyond anyone's control. No one can predict the future.

Third, in risk, you can assign probabilities to situations within a given set of parameters.

In contrast, uncertainty lies beyond the realm of probability and statistics, the Pareto Principle, and even Benoit Mandelbrot’s fractal math.

The crash of 1987 was a 25 Sigma event. The "odds" of a negative 20.5% drop in a single day was about 1 in a trillion.

The only thing worse is the odds of A.I. predicting the next 1 in a trillion crash.

As one writer summarized:

With measurable risk, you can know the odds of outcomes, although you don’t know exactly what will happen in any given case. With uncertainty, you do not even know the odds, and more importantly, you cannot know the odds.

The takeaway?

Bulls say A.I. may one day surpass human investing.

Again, I call B.S.

The A.I. of the future will become just another investment philosophy - like value, momentum, or insider buying.

Sometimes it will work.

Other times it won't.

But, no, A.I. is not the key to unlocking the markets.

And in a world of uncertainty, I predict it never will be.