Alpha Ex Machina
What do healthcare, self-driving cars, voice recognition and search engines have in common with the investment industry? Well, for one thing they all involve analyzing vast swathes of data in order to achieve their objective.
Another is that they, along with countless other areas, are beginning to leverage the power of machine learning to do so.
Learning how to learn
Machine learning was initially developed in the mid-20th century through an academic effort towards constructing artificial intelligence (AI). But it quickly took on a life of its own when its commercial uses began to be exploited.
Put simply, machine learning is a blend of modern computer science and statistics that allows computers to autonomously learn the relationships within a data set. Many of these patterns may be too complicated or obscure for even the most discerning analyst to discover, giving users an edge over traditional methods.
Combined with innovations in big data, the availability of ever larger quantities of specialized information, machine learning is allowing researchers to make inferences that they would have never been able to before, improving our understanding of how the world works and enabling us to make more accurate predictions about the future.
The techniques have been exploited in a broad and growing number of fields. Within healthcare, machine learning is assisting diagnoses. Search engines such as Google use it to optimize search results. Cornell University is using it to analyze whale calls to prevent collisions with ships.
The topic of machine learning is in vogue lately. Scientific and industry figures from Stephen Hawking to Elon Musk recently signed a document warning of the dangers of advancements in the field’s bigger brother, Artificial Intelligence.
Their concerns are dramatically echoed in a spate of Hollywood features such as “Ex Machina,” “Avengers: Age of Ultron” and the recent re-launch of the “Terminator” franchise. All these play on the trope of sentient robots violently turning on their creators.
So is the investment industry in danger of designing its own mechanical demise? Fortunately, the gap between machine learning and true AI remains an extremely wide one, but that hasn’t stopped an increasing number of investment firms and hedge funds from launching computer-driven strategies.
Search for “machine learning” on the Social Science Research Network (SSRN) website, and you will discover that seven of the top 10 most downloaded papers on the topic relate to investment strategies. And the interest is not solely restricted to the academic space.
Companies including Two Sigma, Bridgewater Associates and Renaissance Technologies have all found success in building portfolios that allocate based on quantitative trading suggestions, and a growing number of investment banks are employing machine learning experts in their research teams. For example, such models can be designed to provide asset allocation recommendations, forecasting whether equities will outperform bonds or which equity regions funds should be overweighting.
Sophistication or sophistry?
The possibility of employing machine learning in investment management will likely be met with skepticism from more traditional investors. This suspicion is not unwarranted. It is often quipped that one of the best-selling books on statistics is “How to Lie with Statistics” by Darrell Huff.
The introduction of esoteric machine learning methodologies may make it easier to confuse results and present backtests. When implemented carefully, with a philosophy of scientific rigor and real-world practicality, these new methods are proving to be powerful tools for many facets of investment.
Computer-driven investing is gaining momentum and looks set to transform fund management. While some still draw a divide between the qualitative and the quantitative, it is becoming clearer that, by harnessing the power of machine learning alongside more traditional research skills, decision making can be improved. At least, we believe, until the robots take over.