Revenge of the humans

How ironic. Bloomberg is reporting that most of the new quant strategies adopted by asset managers aren’t generating any, uh, quantifiable positive returns, let alone alpha. Allow me to explain in a moment of schadenfreude.

For years I’ve listened to tedious, unproven projections and inflated rhetoric about the promise of big data, quantamental strategies, roboadvisers, machine learning, web scraping and artificial intelligence.

I’ve been told that humans were passé. I’ve been handed the new quant bible, Homo Deus by Yuval Harari, which argues that human beings are merely algorithms made out of meat, and the best we can hope for is to merge with supercomputers.

Ah, yes, the golden future. Temperamental traders replaced with placid robots, and portfolio managers replaced with data managers that don’t swear, drink or womanize. After all, only a PhD can design an algorithm that web scrapes for how many times Target is searched on Google, and then combines this with credit card records from Visa and compares this to Facebook “likes” and adds a prediction based on photographs from drones circling the parking lots, puts all this alternative data into context with the historical movements of the stock and the indexes generally, and arrives at a nearly infallible prediction of how the stock will move. And it will even learn from its mistakes.

I even heard that one of the new prophets of AI conceded that we should “include humans in the loop” for the time being. How magnanimous. It’s like my laptop telling me that I have to leave the house and move into the little room above the garage.

Many of us anticipated this failure (see “The quantitative tipping point,” below).

First, algorithms only work ceteris paribus, i.e., in a closed system where all other things are equal. But a closed system is never the case. Markets are affected by wars, famines, political posturing, central bank interventions, mass delusions, apocalyptic rhetoric, crypto-currency survivalists and even the weather. No computer can figure all this out.

Second, computers cannot understand symbolic human communication because they lack any social context. The economy is a human construction and it is symbolic and cultural all the way down. There is no mathematical formula lying below the surface waiting to be discovered. No matter how fast an algorithm is, it cannot listen to an investor presentation and decode clues, get an intuitive feel for situations, pick up vibes and develop a feel for what the speaker may be feeling.

Consider the battle over Herbalife (HLF). No matter how good a computer could crunch the numbers, the decisive factor moving the stock price was a monumental clash of egos between Bill Ackman and Carl Icahn. No algorithm can compute pride, narcissism, recklessness, self-promotion and feigned indignation. These same factors are at play in most if not all companies.

The great philosopher Immanuel Kant said, “Out of the crooked timber of humanity, nothing straight was ever created.” This means that as long as human beings are running the show, everything that they touch — from politics to economics to relationships — will be largely irrational, chaotic, indeterminate and contestable. The upper hand will always go to the human being who has studied non-mathematical subjects like history and psychology. It’s true that in closed systems like chess no human can beat a computer, but the superiority of the computer is only made possible by reducing the messiness of the real world to the artificial construct of a game.

Need further proof? Consider waking up to find that the entire market was down because a maniacal dictator halfway around the world sporting a bad haircut is shooting nuclear-capable rockets in our direction. Please tell me what computer, what algorithm, can reduce that to a mathematical formula?

Third, there is a problem of the commons. Once a single firm uses an AI strategy, other firms copy it, and this reduces the spreads and eliminates the edge. If all the computers are learning from each other and their own mistakes, they will eventually reach parity and eliminate any edge for a single fund.

And yet the myth persists even among legends. Dealbreaker reported that Paul Tudor Jones II had cut “15% of the humans” at his hedge fund (as opposed to cutting 15% of the furniture?), and proclaimed, “No man is better than a machine, and no machine is better than a man with a machine.”

Anyone who has watched Jones give a Ted talk knows that he is concerned with corporate justice, alleviating poverty and with improving education. But surely he knows that we’ve had the world’s most powerful computers for decades, and yet these problems persist. To think that computers can solve investment problems but not other human problems is to divide the world artificially, and wrongly.

At bottom, computer worship is simple paganism — a person carves a god from a piece of wood and then claims that the god created him instead of vice versa. We do the exact same thing when we create and then bow down before our own metallic creations.

Corporations are not people. And neither are computers. SIRI is not your friend, and your Japanese sex bot is not your wife. It’s not cutting-edge to stick your own hand inside a puppet and then let the puppet tell you what to do.