Quant Doubts Over a Robot-Led Future Spill Out on a London Stage

(Bloomberg) -- Addressing about 400 aficionados of quant investing, a seasoned hedge-fund adviser asked if anyone in the crowd had seen practical uses for AI that will be recognized as the real deal over the next decade.

“I got my eyes on both of you,” Stuart MacDonald of Bride Valley Partners quipped when just two people raised their hands.

MacDonald was hosting a panel in London on the future of systematic finance at a time when more and more traders attempt to harness artificial intelligence to power rules-based strategies. But the skepticism expressed at Friday’s Quant Conference underscores growing doubts over the disruptive promise of the technology -- one that’s long been touted as the next leg of the tech revolution.

A slew of AI use cases in finance are merely examples of high-level statistics, said Ewan Kirk, chief investment officer of Cantab.

“We wanted Robby the Robot -- and what we got is good movie recommendations,” he said at the gathering, referring to a futuristic icon in the 1956 film ‘Forbidden Planet.’

Machine learning and artificial intelligence cover a broad spectrum of techniques, ranging from complex statistical analysis to mimicking the way neurons in the brain provide layers of learning. For proponents, the technology holds the possibility of uncovering new investing signals, speeding up routine tasks and replacing human traders.

All that is a tantalizing prospect for the finance industry, and especially for quants that have been grappling with underperformance and competition from cheap exchange-traded products.

Yet so far the AI boom hasn’t proved a game-changer. A Eurekahedge index, for example, tracking those using AI and machine learning has returned 6% in the past three years through September, nearly half of the performance of hedge funds overall.

Netflix Example

For Kirk, the movie recommendations are a handy metaphor for why it’s hard to apply AI techniques to finance. The capital markets are so large and complex that getting a clean read from them is a huge challenge.

“Imagine how good Netflix movie recommendations would be if 19 times out of 20 you just watched a random film, and one time out of 20 you actually watched a film you like,” he said. “The recommendations would be terrible.”

Matthew Sargaison, the co-CEO of Man AHL, acknowledged on the same panel that, for all the public interest of the past five years, progress has been slow in uncovering practical uses for AI in day-to-day trading. Still, he was keen to highlight a few positives.

“One of the biggest developments is in the processing of text,” he told the audience. “Using better techniques and investing more heavily in that gives you a significant edge once you get data that are able to be parsed in a number of ways.”

Sargaison said that, while deriving sentiment signals from text initially relied on basic dictionaries, it’s become possible to use more sophisticated techniques to glean information more relevant to markets.

‘Incredibly Hard’

Even at a session about how technology will transform the front office, Matthew Hampson, deputy chief digital officer at Nomura Holdings Inc.’s wholesale division, downplayed the hype.

“We are starting to see broad automation where small low-value activities are starting to be picked up by machines using some of these techniques.” But Hampson added: “It is an incredibly hard problem to deal with client flows on the trading floors in a market-making business and to do automation.”

Even if robots aren’t taking over, things are changing on the trading floor, including at discretionary asset managers, according to Rob Boardman, chief executive officer for Virtu Execution Services for EMEA.

“The leading firms now have employed data scientists and they are looking to reduce implementation costs,” he said. “It’s good for your careers -- they are now hiring quants.”

As Cantab’s Kirk quipped, “AI is always 20 years away.” It’s an old joke in machine-learning circles, but it speaks to the ongoing challenge for quants. AI is undoubtedly the future -- it’s just unclear when it will arrive.

To contact the reporter on this story: Justina Lee in London at jlee1489@bloomberg.net

To contact the editors responsible for this story: Samuel Potter at spotter33@bloomberg.net, Sid Verma

©2019 Bloomberg L.P.