In an age of self-driving cars, ‘robot surgery’ and computers capable of trouncing human players in hugely-complex games such as Chess or Go, it seems obvious to many that the automation of Wall Street, the City of London, Frankfurt and other financial centres must be imminent.
It is assumed that Artificial intelligence (AI) will shortly consign the stock picker and fund manager to the same historical oblivion as the textile weaver.
Now it is quite true that AI has great potential in financial services, of which more in a moment. But in the here and now, the exaggerated expectations of technophile observers have, alas, to be gently let down.
When this happens, the reaction of these observers is frequently to swing to the other extreme, and complain that the financial services industry is failing to take advantage of AI and is stuck in a technological dark age.
At root is a misconception about the application of technology in financial services. Machine learning, while a less catchy concept than AI, has been used for a long time. After all, a simple linear regression in an Excel spread-sheet is an example of machine learning, albeit of a basic type.
No, this is not the financial-services equivalent of the self-driving car. Nor is it meant to be. Machine learning is a less intriguing name than AI, but it is widely and profitably used.
Furthermore, advanced machine learning can act as a powerful partner for human managers. It can generate predictions, and then turn those predictions into a strategy. It can pick long/short positions more swiftly than a manager, but in all cases human supervision is needed to ‘sense check’ the proposed course of action.
The machine should not be allowed to operate without oversight.
Here is one, very pertinent, example. There was, as we all know, a sharp fall in the external value of sterling once the result of the UK’s June 23 2016 referendum on membership of the European Union became known. A machine learning program would study the huge amount of historical data at its disposal and note at once that every such sharp decline in the value of the pound in the past, such as 1976, 1985 and 1992, when Britain left the euro Exchange Rate Mechanism, had been followed by a ‘bounce’.
This would have seemed, from the machine learning perspective, an obvious buying opportunity. It would take a human manager to sense that ‘this time it’s different’, that the referendum result had taken us into uncharted waters, and that it would be prudent not to assume a strong recovery in sterling.
Such a manager, of course, would have been vindicated. Sterling may be off its immediate post-referendum lows, but at the time of writing, as 2017 comes to a close, it remains well below its pre-June 23 levels.
In short, automation is not going to replace humans any time soon. But that is not to decry the value of developments in ‘big data’ that can greatly enhance machine learning improve outcomes for clients, even if it’s not as evocative as ‘robot managers’.
Indeed, with the proper use of data, firms can both obtain a better understanding of their clients and help to educate those same clients in different trading styles. It should come as no surprise to learn that no asset manager is spending less on data than they were five years ago, and most are spending more.
In part, this reflects the value placed by the asset-management industry on such data. In part, it reflects the realisation by the vendors of the worth of their product, with the result that prices have risen sharply.
A parallel challenge facing managers in the industry is the need constantly to increase the skill set of the people in their teams, with special reference to technology. More and more, they will be looking for multi-skilled people. For example, a trader who is also a programmer.
When considering any developments in this area, the key tests will always be, first, the return on what can be a hefty investment and, second, the benefit to the clients who are, after all, what the entire exercise is about.
People get excited about the sci-fi aspects of technology in financial services, as they do elsewhere. They feel let down when told that the robot fund manager remains very much on the drawing board.
However, they ought to take heart from what are exciting and innovative developments in the use of data analytics, making more powerful still the partnership between human managers and machines.
The aim of the exercise is to generate the best returns for clients and, for now, that means using the technology as a tool to be wielded, ultimately, by a human manager.
Charles Ellis is a trader and quantitative strategist at Mediolanum Asset Management.
Keep up to date with all the latest personal finance news and investment tips by signing up to our newsletter. Email subscribers will also receive a free print copy of Money Observer magazine.