Wall Street traders are getting crushed by AI

An international team of researchers from the School of Business and Economics at Germany's Friedrich-Alexander-Universität Erlangen-Nürnberg have managed to show that over the past thirty years artificial intelligence (AI) has been making a killing on the stock market.

That's why today, as the attendees of the Battle of the Quants event I spoke at recently already know, an increasing number of funds, from hot innovative "Quant" start ups such as Numerai and Quantopian, through to the world's largest such as Bridgewater Associates and Renaissance Technologies, are using AI as their primary investment weapon of choice.

Scoring above average returns on the stock market is notoriously hard, and I can tell you for nothing that my dart board and darts approach just plain sucks, but maybe I should take solace in the fact that apparently even the best brains on Wall Street find it hard to pick winners reliably. However, if Krauss' research, which was published on Science Direct, is to be believed then the problem that me and my Wall Street colleagues share is that, well, we’re human. And ouch – that’s got to smart.

However, the AI stock pickers didn't just beat the market. They annihilated it.

One model, created by one of the oldest quants, returned a 73 percent annual return from 1992 to 2015, and that’s after accounting for transaction costs. Compare that to a real market return of 9 percent and annihilate might be too weak a word. The team also found that, unsurprisingly perhaps, that gains were particularly high during times of high market turmoil - the moments when human investors are often overcome by greed or fear and make emotional, and sometimes irrational, decisions – something that AI’s aren’t “afflicted” by.

For example, in 2008, when the global financial meltdown cost many investors the shirts off their backs another AI system notched a breath taking 681 percent return, and in 2000, when the tech bubble burst, that same AI returned 545 percent.

"Quantitative algorithms turned out to be particularly effective at such times of high volatility, when emotions dominate the markets," said Dr. Christopher Krauss who was the lead author of the study.

Increasingly AI is encroaching on territory that we, as humans, thought we our domain and our domain alone, and, furthermore with every new announcement it seems that they’re not just moving in – they’re taking over and besting us. And it doesn’t look like finance is very different.

The results of Krauss' study suggest AI, which is already being used to run the world’s first fully autonomous hedge fund, Aidyia, might one day dominate the world of financial strategy making, too - a state of affairs that would have hard to predict ramifications, which isn’t made any easier by the fact that AI is already seen as a “black box” technology.

While most of the world’s equity trading is already automated and run by AI, for example over at Goldman Sachs where they managed to reduce six hundred equity traders down to just three, the automation of finance is an increasing trend, and the proponents of unleashing advanced AI onto the financial markets argue that it represents a new level of sophistication and flexibility, along with unprecedented firepower. Where firepower means earnings potential.

Most of today’s state of the art AI’s can learn from past experiences, something that itself is about to take a quantum leap forwards with a new AI breakthrough that all but eliminates what experts called “catastrophic forgetting,” and adjust their trading strategies instead of following rote, relatively simple trading algorithms. And I won’t even mention the fact that AI’s are now at the point where AI’s can create new AI’s and code, or evolve, their own programs – two developments that haven’t hit Wall Street yet but that will one day soon.

In the meantime though, with huge amounts of money at stake it’s inevitable that some investors are already turning AI loose on real world markets.

Aidyia’s autonomous AI, for example, can read news in multiple languages, analyse reams of economic data, identify obscure patterns, make predictions about market trends, and then invest accordingly, and, according to its creators “if we all die, it would keep trading."

Other companies bringing AI to bear on financial markets include Sentient Technologies in San Francisco, and Rebellion Research in New York, but unlike Dr. Krauss' public study much of what is being developed and tested by real world markets remains wrapped under a veil of corporate secrecy for fear of letting competitors gain an advantage.

In other words, investors who might now be making a killing in financial markets using AI may not want the world to know too much about it, because the more widely it’s dispersed, the less advantage it brings to those who have it, and that effect appeared to show up in Dr. Krauss' study as returns of the researchers' AI investment methods declined after the year 2001 as computer based trading became more widespread, suggesting less room to exploit market inefficiencies.

"'In the latter years of the study, profitability fell," Krauss said, "we assume that this decline was driven by the rising influence of AI in modern trading."

As for myself, however, I’m ditching my pen and paper and building myself financial Skynet. Sorry, I meant I’m going to build myself a carefully crafted machine learning system that helps me pick penny stocks… and if you want to know more then why not come and meet the experts at the next Battle of the Quants in New York? Just don’t forget to bring your own AI.