Instability: When Machines Take Over

Behind the broad, swift market slide of 2018 is an underlying new reality: Roughly 85% of all trading is on autopilot—controlled by machines, models, or passive investing formulas, creating an unprecedented trading herd that moves in unison and is blazingly fast. Behind the Market Swoon: The Herdlike Behavior of Computerized Trading, The Wall Street Journal

As an engineer, I know a few things about feedback loops. The world is full of feedback loops. I design electronic ones, but the financial markets are full of them too. When a bunch of people build automatic control loops around some crucial piece of infrastructure (like the stock market) and gradually turn up the gain to maximize profits (as we know they will), the system will eventually go unstable and oscillate. You can model that. It looks like someone learning to drive a car while spotting the road directly in front of the hood, overcorrecting for every perceived error.

Meanwhile, to the great detriment of humanity, many bright minds–including many of my college classmates–are engaged in high-tech R&D to advance the state of the art in automated “investing.” They have built new algorithms, better supercomputers, broken new ground in programming languages and big-data analysis in the name of automated trading. Millions of dollars have been invested in further gaming the system, like by locating datacenters in places that optimize communications latency at sub-millisecond levels.

Does any of this work serve the real world? Of course not. None of it creates lasting value–just money. Automated trading exists to siphon money out of the real economy and into the hands of the financial industry. Appreciate for a moment that it’s not really fair to call these schemes “investing” if they are automated and high speed. Investing implies that you are making a long-term bet on something, not responding to a tiny signal.

I’ve met these people and their companies at career fairs where we compete for young talent, and it feels bleakly hopeless and strange to me that nobody seems to question the ethics of it. Of course it’s technically demanding and academically interesting work, but so is building weapons of mass destruction. Except instead of building machines to kill people, you get to build machines that gamble with people’s retirement savings.

If I were king of the world, I would fix this problem quickly and easily without violating any of the core principles of capitalism: by introducing a highly random and arbitrary (but fairly short) delay on every trade, automated or human. There is no real reason that trading needs to be fast. Overnight, automated computerized trading would be wiped out, but fairly so. The big banks siphoning money out of the system through high-speed financial arbitrage would lose a potent source of income, but this consequence would pose no detriment to the actual (productive) economy.

Think about it.

Leave a Reply