In March, Japan’s ¥158trn (€1.2trn) Government Pension Investment Fund (GPIF) and Sony Computer Science Laboratories (SCSL) developed a deep learning prototype “to detect the investment style of managers”. The resulting AI technology had been programmed to differentiate between fund managers’ styles and associated drifts, “enabling evidence-based, prompt analysis of investment styles”.
GPIF and SCLS are not the only investors (nor indeed the first) to apply deep learning to analyse the investment strategies and styles of funds. At Fidelity International’s annual Investment Conference in October, it was interesting to see UK-based Irithmics showcasing some of the current capabilities of deep learning being applied to fund strategies and styles already being used, including by institutional investors for fund screening, selection and monitoring.
Fundamentally, fund selection is an investor’s effort to identify a product satisfying their investment objectives. This balances risk and reward together with other aspects, from tax and regulatory considerations, to the intricacies of investment strategy, plus how profits are generated and risks mitigated; this constitutes the investment strategy.
While it is easy to calculate, standardise and compare fund performance, it is significantly more difficult to objectively evaluate the strategies which generate the performance of funds.
New insights now exist which can support investors in their efforts to assess strategies and to monitor them once selected: powerful new insights now exist that can support the evaluation of the opportunities and risks of a fund, helping to identify and manage the risks of a portfolio of funds.
It is this ability of AI to enhance the understanding of a fund’s strategy which will transform the way funds are screened and selected in the future.
As investors turn to technology to help differentiate funds they are able to survey the rich universe of funds in new ways, select the more innovative strategies and developing more diversified and resilient portfolios.The relationship between investors and managers will inevitably evolve in this new climate. Investors will differentiate between funds based, as SCLS describes, on “evidence-based, prompt analysis of investment styles”.
The existing capability of AI to screen and differentiate funds, for example as demonstrated by Irithmics’ deep learning capabilities, highlight the potential impact of the technology on fund selection and the enhanced insights available to investors.
The impact of AI on fund screening and selection is likely to be felt beyond investors however. Armed with AI-generated assessments of a fund, its strategy, opportunities and risks, investors will look to managers to provide more detailed explanations. Managers will come under increased pressure to explain very specific aspects of the strategy – their inability to adequately do so will undoubtedly colour an investors’ view of the fund.
How AI has classified a fund, how it describes the strategy, identifies opportunities and risks will become equally important to managers as they seek to leverage the machine's assessment of the fund to make stronger investment cases for their funds.
It is clear why large fund investors like GPIF are turning to technology to enhance their fund screening, selection and monitoring. The ability to differentiate between fund strategies is a key differentiator to building well diversified and performing portfolios of funds.
The ability to better comprehend a fund’s strategy will lead to better decisions about risks and how to effectively manage them.
Irithmics recently began prototyping deep learning technology to assist with the mitigation of the downside risks associated with portfolios of funds. It is this combination of technology better profiling and screening of funds, combined with the ability to proactively use the advantages of machine intelligence, which will transform the way investors approach funds in the future.
Stuart Fieldhouse serves as a Director at Hawksmoor Partners