If there is one industry that would greatly benefit from the genuine adoption of artificial intelligence (AI), it is investment management.
Investing, in its simplest form, is about predicting the future — the future value of a company, index, currency, country, or relationship. Investment predictions are made like all predictions: combining information with a model or method. And we assume that we have some advantage, some edge, which allows us to make predictions that are more likely to be right than wrong. Otherwise we could flip a coin.
This edge, which is the source of the ever-elusive alpha, is the result of either superior information or superior decision-making methods.
However, the failure of managers to consistently generate the promised returns reveals that managers generally lack whatever edge they seemed to have had in the past.
This is because active managers generally use the same information (price, economic and financial data) and the same methods (some iteration of multifactorial linear regression laden with a heavy dose of mean variance optimisation) to make their investment decisions.
The homogeneity of data and methods has caused even the historically best active managers such as Warren Buffett to struggle.
Fortunately, there is AI, whose raison d’être is to make predictions by applying powerful computational methods to large, often complex data sets. Advanced AI such as deep learning and deep reinforcement learning has been able to make predictions with superhuman accuracy.
Yet as a recent CFA report showed, only a handful have realised the promise of AI and directly incorporated AI into their investment processes.
While small in number, their influence is great and their adoption is forcing their less innovative peers to publicly claim that they, too, have adopted AI.
However, their claims are disingenuous and typically distil to one of three types of posturing:
-
Introducing a non-traditional data set (eg, credit card data) into their existing traditional, non-AI investment processes;
-
Window-dressing their staff by adding a computer scientist or two that they bring out for client meetings;
-
Simply misappropriating the term “machine learning” to include traditional quantitative processes such as cluster analysis or linear regression.
These pretenders diligently work to lower client expectations of the power and application of AI by asserting, without evidence, that humans “will always be a necessary component in the investment process”.
This anthropogenic requirement not only artificially fixes clients’ expectations; it excludes the use of the equivalent of “Thor’s golden hammer” — deep learning and reinforcement learning — in asset management.
This artificial requirement is debilitating because deep learning and reinforcement learning are precisely more powerful than classical machine learning and, in more and more cases, human intelligence, because they are constructed and provide an output without the use of human data or input.
So if AI could help managers achieve the edge they so desperately need (and with it, the fees and clients they desire), why has the rate of adoption been so slow? And why have managers systematically erected artificial barriers to the use of deep learning and reinforcement learning? Because managers recognise that true adoption requires true disruption.
AI not only fundamentally changes a manager’s investment process and investment worldview, but it also radically transforms the entire business — it requires new talent, new data, a different R&D process, a new management structure, a new brand and, perhaps most importantly, new funds to pay for this multimillion-dollar endeavour. All these changes come with absolutely no certainty of achieving any net benefit or an ROI.
Given these circumstances, too many incumbent managers cleave to the status quo. After all, in spite of mediocre investment performance and misaligned fee structures, many active managers have experienced only modest client redemptions and fee compression, leaving them to enjoy profit margins greater than 30 per cent.
By choosing the status quo over disruption and incrementalism over innovation, managers are placing their commercial and personal interests ahead of their clients’ interests because the status quo simply cannot provide their clients with the investment outcomes they seek.
They make this choice in spite of evidence that the majority of institutional clients consider it critically important or important that managers are already using AI in their investment processes and that they use AI themselves to disintermediate the investment decision-making process.
Consider that the world’s largest pension fund, Japan’s Government Pension Investment Fund, has partnered with Sony Computer Science Laboratories to internally develop and deploy a deep-learning-based manager evaluation application.
Choosing the status quo over disruption will certainly prove fatal for managers; the question is, how long can they live on their current life support systems?
Angelo Calvello is co-founder of Rosetta Analytics
