Many fund houses are divided upon their views of artificial intelligence (AI) funds. Some see them as an existential challenge. Others, a major opportunity to cut costs and improve fund investing accuracy.
To date, several funds to be formed that are more sophisticated than basic passive fund investors. But they all failed to anticipate the market volatility of early 2018.
Florian Spiegl, co-chair of the artificial intelligence committee of the FinTech Association of Hong Kong, believes the recent underperformance of machine learning-directed funds suggests they were designed to operate in the one-directional markets that prevailed across most of the past two years. When those conditions shifted, they were not well-equipped to adapt.
“This hints that strategies which are entirely relying on contemporary AI approaches have not been through sufficient training in different market cycles, despite the availability of historical data,” he said.
Similarly, Mohammad Hassan, head of hedge fund research at Eurekahedge, believes it will take time for AI hedge funds to adapt to market cycles.
“I’m not sure there’s any AI fund that could predict that [US president Donald] Trump was going to slap tariffs on US steel imports or the prospect of talks with North Korea. These factors are hard to predict and the AI funds are still adapting.”
For AI sceptics, such mistakes underline the limitations of the technology. Larry Wang, head of marketing at CSOP Asset Management in Hong Kong, says his firm is not hurrying to launch AI-managed funds. He believes the technology is not yet advanced to make the case for funds that purely manage investing via machines.
Wang says AI funds depend heavily on macroeconomic data and computation, but this doesn’t always provide the full picture. Additionally, he contends that much more work needs to be done on how to correct AI technology. “Where there are errors occurring in decisions made by AI technology, humans still struggle to identify the reasoning and logic behind it.”
The AI ecosystem is also immature, he argued. “The core intellectual property is still in the hands of the industry leaders. There is a disconnect among the big players that will stunt the process of AI development.”
Even the optimistic Spiegl concedes that in its current state, “machine learning has still more of an augmenting effect on the investment process, less a displacing one”.
But advocates of AI are enthusiastic about the future. Hassan notes that AI funds growing out of traditional quantitative systematic strategies have already demonstrated their capabilities.
“Right now, we are in a transition phase between the CTAs (commodity trading adviser funds, which typically use futures contracts to invest), which have suffered over the last couple of years and posted their biggest loss ever in February, and the new wave of AI funds,” he predicted.
On a five-year annualised basis, the average CTA returned 2.63% a year while and the average AI hedge fund returned 10.71%, according to Eurekahedge. This is one of the push factors causing asset managers to turn towards AI.
Hassan believes CTAs have run their course, and that AI funds will refine their ideas and increasingly become the focus of this type of investment strategy.
Frederick Chu, head of exchange-traded funds at China AMC, is also enthusiastic. “AI is going to be a driving force for the investment management industry,” he declared. China AMC’s Hong Kong division is the first Asian asset manager to introduce an AI product into the region—an A-Share multi-factor strategy, launched in mid-March. The strategy is based on a machine-learning stock-picking algorithm, which applies a smart beta-like multi-factor approach, making changes for different market cycles.
These are just opening examples. Over the coming decade, many more AI-supported funds are likely to enter the fray, In 10 years “I think we might see virtually all discretionary traders lose their jobs—we’ll need one-tenth of the amount of trader positions that we have now,” Chakravorty said.
CALCULATION VS. INTUITION
Will the machines ever evolve enough to completely take over from human CIOs? David Asplin, managing director of global business development at Brisbane-based Queensland Investment Corporation, doesn’t think so.
“Not in terms of the decision-making. There’s absolutely a role for machines to play, but the human interface is also very important,” he said.
Spiegl paints a more nuanced picture. He argued future AI funds will become more sophisticated as they learn to deal with different market cycles, which means they will become better quantitative fund operators.
“Machine learning continues to gradually take over the entire investment process of quantitative funds,” he said. “Machines will ultimately deliver higher speed, quality and precision compared to human investment managers, across different cyclical phases in markets.”
For him, two interesting questions remain: “Which role will deeply human qualities such as intuition continue to play in a highly automated investment process and will we as humans set limits to the use of AI in our financial markets, just like we did with high-frequency trading?”
Asset owners have more prosaic concerns. Hiromichi Mizuno, chief investment officer of Government Pension Investment Fund, the world's largest pension fund, is ultimately agnostic as to whether funds are run by AI or human beings, provided they meet GPIF’s long-term goals.
“I’m not sure that AI can replace a human manager, but that’s the asset manager’s job to decide,” he said. “If you think you had better replace your team with the AI you should do it to give us the best product.”
Richard Morrow contributed to this story.
Click here to read the first part of this feature from our April/May edition on the rise of artificial intelligence funds.