LIMA MBEU: Expert Opinions: Edition March / May 2020

image_pdfimage_print

The role of machines in investing

Ndina Rabali, chief investment officer at Lima Mbeu Investment Managers, provides a perspective.

Artificial intelligence, machine learning, deep neural networks. These terms cause apprehension amongst asset managers and fund trustees. Gary Smith, a professor at Pomona College in California, puts them into context:

Computers can input, process and output enormous amounts of information at great speeds. Computers are relentlessly consistent. It is therefore tempting to think that computers are smarter than humans because they do some very difficult tasks better than humans.

Computers may be more efficient at discovering patterns, but they are still incapable of assessing whether those patterns are useful or merely coincidental. Only humans can make this assessment.

Computers do so many things well, but this does not mean that they are better investors than humans. This is why we prefer a Quantamental investment approach that integrates quantitative techniques with human judgement in building investment portfolios.

A simple example

Let us assume that someone wants to buy a house and that this house must satisfy a list of requirements, including price, size, location, style and functionality. What process would they typically follow?

Firstly, they would identify a list of potential houses to look at in their preferred locations by talking to estate agents, going through newspaper adverts, or trolling through various websites. They would also evaluate crime statistics for the area, drive around to observe traffic patterns and attend multiple show days. This is the traditional approach.

Or they could use artificial intelligence or machine learning tools to assess traffic patterns, crime statistics, size, functionality all at the click of a button. This would save time, money and a lot of unnecessary effort.

Use of a data-mining tool, to generate a list of potential houses to buy, not only contributes to improved efficiency but also removes bias in the process. For example, when driving around to evaluate traffic patterns, it could be that there was load-shedding on a particular day leading to a one-sided, biased view against a specific location. But a datamining tool can evaluate traffic patterns more accurately – without wasting petrol.

Secondly, they would conduct thorough due diligence. This involves inspecting the list of houses they have identified for faults that may not be immediately observable. In this case, technology may be of little use because human judgement will be required to identify significant or coincidental issues.

For example, assessing how a fresh lick of paint has covered the cracks on a wall would be too difficult for a data-mining tool. The point is that the most efficient process is one that combines the use of both human judgement and computing power. Some tasks are best left to a computer, and some functions should retain an element of human judgement.

For optimal results

In the same way, Quantamental investing aims to harness the best of both worlds to deliver superior returns for investors. They may use a data-mining tool or an advanced factor model to generate an unbiased list of investment ideas. However, human judgement still has a role to play in assessing ideas before they are implemented in a portfolio.

Investors and trustees are right to be apprehensive if they choose to continue ignoring the benefits that they can get from the use of advanced statistical techniques. This is why we believe that future success in asset management will require technological efficiency in the processing of information through expert systems. They assist in forming, controlling and implementing portfolios.

www.limambeu.co.za