Fischer explained that machine learning, when machines have the ability to improve performance over time without human intervention, would need to “see thousands of gains” from trading positions in order to have adequate data to learn, which in itself is a limitation.
Also limited is “deep learning” — essentially teaching computers to learn by example — is very difficult because of the lack of data in financial markets.
For instance, Fischer said that artificial intelligence cannot analyse corporate governance changes in Japan. “That is more than two or three people making decisions, and machines can’t yet figure out the probability of one person making a decision.”
The same goes for analysing the merger of two huge companies. “There are lots of different decision points. You have to read through different people’s minds.”
George Long, chairman and chief investment officer at Lim Advisors, added that machines cannot analyse some markets that have huge inefficiencies.
Long acknowledged that the use of big data — comprehensive data collection and analysis far beyond traditional methods — may work better in certain markets, particularly in developed markets, because of information availability. But big data strategies do not work satisfactorily in other markets such as in Asia.
“There are a lot of embryonic markets like Myanmar, where they are just starting up a stock market. So how do you set up your AI for that? Even in China, there is a lot of big data, but there is friction caused by cross-border restrictions and capital controls.”
Also mentioning Japan, Long said that managers cannot set up a programme to analyse how Japanese companies think.
In spite of the limitations of AI strategies, managers see growing demand for smart beta and factor investing products, which are largely run by machines.
“For our investors now, what they like is the predictability of machine-managed money, but they also like the outperformance that the humans still bring,” Murray Steel, chief operating officer for Asia-Pacific at Man Group, said during the discussion. “So there is a role for both for a very long time.”
In addition, there are some tasks that machines can do better than human beings alone. For example, they are faster in gathering and analysing company earnings revisions and outlooks, Oasis’ Fischer said.
Besides the impact of machines in the hedge fund industry, the panelists also discussed whether succession planning is relevant in the industry and gave views on key man risk — when investors rely too much on one person such as the fund manager.
Man Group’s Steel believes that hedge fund firms should have in place a succession plan.
“This is something that our investors are really focused on. They don’t want to see a key person risk.”
He explained that investors do not like the idea of following a “star manager” or a star analyst. “They value the company, the brand and the infrastructure that comes with that.”
Oasis’ Fischer shares the same sentiment: “It is part of good governance. You want good governance in companies, you would want good governance internally.”
However, Lim Advisors’ Long believes key man risk is misunderstood. When investors buy into a hedge fund, they are buying a manager rather than a platform, he said.