Investment funds are increasingly using artificial intelligence to analyse data and even select stocks, though most agree it is essential humans retain oversight.
There is nothing new about ‘black box’ funds, investing in quantitatively-managed stocks selected by machines. In fact the career of one of their pioneers, Barr Rosenberg, was prematurely ended by US regulators back in 2011 following a programming error.
But the latest industry focus on artificial intelligence (AI), has put funds driven by new technologies firmly back in the spotlight.
The prominent AIEQ, which claimed to be the first public fund driven by AI, was launched on October 18, 2017. Utilising machine learning technologies, the fund actively selects stocks for its portfolio. It quickly gained popularity, raising more than $70m within a few weeks of inception.
Since then, more funds have jumped on the bandwagon, incorporating AI tools or big data into their strategies. Bridgewater Associates launched a fund in June, raising close to $2bn, with a portfolio strategy predominantly using AI to make investment decisions.
In a 2022 research paper titled ‘Do AI-powered mutual funds perform better?’ published on ScienceDirect, academics came to the conclusion that AI-powered mutual funds show superior stock selection capability and outperform their human-managed peers.
Alternative data goes mainstream
In the era of AI and big data, where ChatGPT has grown faster than TikTok and Instagram, portfolio managers believe there are many positives.
“Relying solely on widely available, mainstream financial data can limit investment perspectives, as this information has likely already been absorbed by the market,” says Laurent Joué, senior quantitative portfolio manager at Lombard Odier Investment Managers (LOIM) in Geneva. “To gain a competitive edge, it is crucial to diversify beyond these traditional data sources.”
LOIM recently launched its DataEdge strategy, using big data on consumer trends to identify “earnings surprises” and capture alpha potential.
“Leveraging alternative data sources can help boost the accuracy of forecasting key performance indicators and predicting earnings surprises,” believes Mr Joué.
The “nowcasting” capability, where real-time data signals are used to make more timely assessments, he says, can provide a significant advantage over traditional, “backward-looking” approaches reliant on lagging financial statements. The future of data-driven investments looks promising, according to Mr Joué.
“The volume, variety, and velocity of alternative data sources will continue to expand rapidly, driven by technological advancements and the increasing digitisation of the global economy,” he says.
“This will provide investment managers with a vastly greater pool of alternative datasets to potentially extract valuable insights from,” he believes. He adds that alternative data will become a key competitive edge.
There are, however, barriers to progress, according to Mr Joué. These include cost of data acquisition, expertise and knowledge, and the need for time to build analytical frameworks.
There is also a personnel problem. “Despite the rising amount of alternative data available, there are still relatively few quant specialists focused on this space,” says Mr Joué.
“Many fundamental analysts are incorporating alternative data as part of a broader ‘mosaic theory’ approach to investment decision-making, rather than developing dedicated quantitative strategies,” he warns.
Black boxes banished
Although use of AI in asset management has advanced considerably in recent years, it is still primarily utilised by large research firms and fintech companies, rather than the fund houses themselves. But this could be about to change.
“We expect AI to gain much wider adoption in stock selection,” says David Wright, co-head of Quest, a strategy deploying AI at Pictet Asset Management, also in Geneva. “With AI, there is always something more to discover or even test. For example, we are currently looking at using deep neural networks in our strategy, and we’re testing other features, such as news articles, to see if we can extract signals,” he explains.
But transparency and understanding of AI-driven investment strategies are integral for Mr Wright. “Employing AI-based models in investment is fraught with complexity and risk,” he says. “The assumption that investors can accept AI as a ‘black box’ no longer holds.”
There is now a need to offer investors a breakdown of positions, risks, and performance of the underlying inputs, he believes. “Transparency is crucial,” says Mr Wright while citing how the Pictet team “deeply” investigated this in a paper called ‘Performance attribution of machine learning methods for stock returns prediction’, published in The Journal of Finance and Data Science. “This and ongoing research allow us to deploy a ‘crystal box’ not a ‘black box’,” he says.
Geeks on the loose
Regulators are also finally starting to notice the importance of AI, says Paweł Skrzypek, co-founder of the AI-driven Omphalos Fund. “Seven years ago, nobody was talking about deregulating AI. It was rather something for the scientists, the geeks, and so on,” he says. “But now, especially in Europe, the AI arc is very serious, and there are very concrete regulations.”
In June, the Omphalos fund upgraded to its 3.0 version, which it says will enhance risk-return profiles, speed up development of new trading strategies and improve overall risk management and reporting.
"This upgrade will further solidify our technology lead in the global asset management industry and should help us deliver even better performance for our clients," says Mr Skrzypek.
Human factor
But despite the fast-improving technology, most portfolio managers agree that the human should never be out of the equation. “You always want to have a human in the loop,” says Mike Conover, co-founder and chief executive of Brightwave, which recently secured $6mn in seed funding to launch an AI-powered financial research assistant. “We should make sure that humans are reviewing recommendations and that humans are the ones making final decisions,” he adds.
He believes that while AI tools are powerful, they are not a substitute for good judgment.
