By Beverly Chandler, AlphaQ.world
Dr Yasin Rosowsky is Head of Artificial Intelligence Research at Arabesque, the ESG asset management firm that uses machine learning and big data to evaluate the performance and sustainability of listed companies.
Arabesque featured in an interview in AlphaQ with its founder and CEO, Omar Selim, back in 2015. At the time, he spoke about the firm’s proprietary technology Arabesque S-Ray, a tool that allows investors to monitor the sustainability performance of approximately 7,000 listed corporations worldwide.
By leveraging machine learning and big data, S-Ray’s technology systematically combines over 200 ESG metrics with news signals from over 50,000 sources across 15 languages enabling greater transparency into corporate behaviour and management.
“S-Ray takes the inspiration of its name from the impact the X-Ray had on medicine”, says Rosowsky. “It’s our way of looking beneath the surface of a company, and analysing its non-financial information.”
Arabesque S-Ray is the first tool of its kind to rate companies on the normative principles of the United Nations Global Compact, the world’s largest corporate sustainability initiative established by the late UN Secretary General Kofi Annan in 2000 with Georg Kell, who is now Arabesque’s Chairman.
One of the main features of S-Ray is that the tool works in almost real time, updating scores and incorporating news-based information through over 30,000 media articles daily.
“Our mission as a firm has always been to mainstream sustainable investment, so we took a decision two years ago to make our S-Ray technology available to the market. We opened it up,” Rosowsky says. “We now provide the scores to other institutional investors and managers who want aggregated, streamlined ESG data at the touch of a button, and to corporations who want to integrate sustainability information into their decision-making processes.”
The firm licenses its S-Ray data scores to a wide range of clients across the global financial landscape, including State Street, BNY Mellon, the Japanese Government Pension Fund and FactSet.
“The integration of sustainability across all of our investment processes is in part aimed at minimising tail risk events,” Rosowsky says. “It’s not necessarily a generator of short term price behaviour, but can be a good indicator of issues within a company and long-term risk-adjusted returns. Some of the biggest corporate scandals in recent years have been picked up by S-Ray, therefore eliminating those companies automatically from our investment universe.”
Another key component of Arabesque’s flagship Systematic strategy is a sophisticated portfolio optimisation method where 1,600 signals per stock are analysed to determine its strength, together with market momentum. A risk management system is built in, which allocates between cash and equity to balance exposure to fluctuations in the market.
“The quantitative, rules-based strategy looks at momentum, with our technology able to extract a good indication of where the trends lie. The risk management system helps us to minimise drawdowns,” he says.
The long-only Arabesque Systematic, a UCITS fund, has returned +10.22 per cent per annum since inception in 2014, and has been awarded a 5-star Morningstar Rating. Last year, the firm also launched Arabesque Systematic USA, a 40 Act fund that applies the same investment strategy to listed US equities.
The last couple of years have seen Arabesque working on building a new artificial intelligence (AI) engine, with the aim of launching a new fund based on that engine.
“At its core, it is a massively distributed network, which connects many different data sources with algorithmic models and machine learning algorithms. Its focus is to answer a single question: ‘should we invest in this stock or not?’ We ask our engine to look at the vast array of complex relationships exhibited within the financial markets and come up with investment decisions.”
The network has over a billion nodes, and engrained within each node are fundamental data, market data, analysts’ reports, and price information. All nodes are connected and are constantly processing.
“It’s computing in six different data centres around Europe to come up with the investment decisions,” Rosowsky explains. “Currently, this represents 1,000 machines working concurrently but we have capacity for 10,000 plus due to its seamlessly scalable architecture.”
“This is new,” Rosowsky says. “It is an evolving and adapting system. Our long-term goal is general AI.”
Narrow AI is the type of machine learning algorithms that solve a very specific problem, but general AI is what many people first think of when you say ‘artificial intelligence’, with computers approaching human intelligence.
“Many people are now embracing AI and entrusting it in their own lives, whether it is in transportation, security or infrastructure, or at home through something like Alexa or Google Home. And we are seeing the beginnings of this now in the investment world too.”