Digital Assets Report


Like this article?

Sign up to our free newsletter

“The stars of the show”: What next for humans in the hedge fund machine learning revolution?

Related Topics

This year’s hedgeweekLIVE Technology Summit opened with an in-depth examination of future trends in hedge fund technology, with panellists exploring how managers are ramping up their use of natural language processing, AI and machine learning tools, and improving efficiency as data consumption grows and skillsets evolve.

The session heard how, historically, quantitative hedge fund models were built around market data, structural data and numerical data. But now an “enormous” new field of text data is available to be mined and to potentially find sources of investment decision-making guidance.

“This is a major evolution; a tectonic shift in terms of the data we have access to,” said Giselle Comissiong, director and head of brand and communications at LFIS Capital, outlining some of the strategic partnerships and initiatives her firm is working on, particularly in areas such as ESG.

“One of the key challenges though is not to put too much confidence in it at this stage,” she added. “It’s really at the beginning rather than being at the middle or the end of being explored and tested.”

Mark Brubaker, chief technology officer at Point72, sketched out the ways in which his firm, which spans both discretionary and quantitative investing, looks to combine and collaborate across the two spectrums.

“While technology is advancing, and we can do some amazing things, there is no replacement for the grey matter and the business experience that is in the portfolio managers’ heads,” Brubaker noted.

Underlining this point, he suggested that data processing can often prove a time- and energy-intensive activity, and may require further augmentation on the part of the firm. “Very rarely is there something that’s out-of-the-box and adds value. In many cases, where that does exist, then the alpha is already gone.”

He added: “It really is a hybrid of data technology, proprietary secret sauce, as well as – most importantly – the portfolio manager themselves and what they do.”

Looking ahead, speakers also gauged how technology and data requirements might drive evolution within hedge fund businesses over the next 10 years, and considered what new expertise and roles firms may require as the relationship between humans and machines gets ever closer. 

Against that backdrop, the session also heard how code blips, flash crashes and algorithms going awry could also throw up dangerous unintended consequences and spikes in volatility for hedge funds and investment managers further down the line.

Kimberly Durland, senior vice president at Arcesium, said the industry can expect to see more coders, more data scientists and more computer scientists in the coming decade. “But on the flipside we’ll see some democratisation of data and its usage in the industry, so we’ll see more low-code and no-code solutions.”

Brubaker pointed to Point72’s business and data training academy, explaining how the firm “believes strongly in growing internal talent” and forging a collaborative spirit and teamwork between data scientists and portfolio managers.

Paris-based LFIS hosts an annual hackathon around machine learning, and Comissiong said the firm also reaching out to the PhD student community, noting how French universities have a long, rich engineering heritage to tap into. “I.T. is sexy again,” she noted. “These people are almost the stars of the show from a quantitative asset management perspective.”

On the potential “unknown unknowns” as machines become integral to investment processes, Matt Angrist, director, financial services at AlphaSense, observed: “We want to provide human beings with tools that makes them more efficient and smarter and better at what they are doing. Everything we are building is more around how we improve their workflow.”

Durland acknowledged that until humans have a better ability to understand the drivers of more the opaque methods and deep learning, there is an “inherent risk” that the computer can take unintentional and unforeseen risks. “The human element is not going away,” she said.

Comissiong stated: “Because of the size and speed and capacity of the automated trading today, this could become catastrophic. We’ve multiplied the power by so much, if the slightest thing goes wrong you risk a major market event. The flash crashes have showed us that.”

Like this article? Sign up to our free newsletter

Most Popular

Further Reading