Remember me

Register  |   Lost password?

The Trading Mesh

Event report: AI and Deep Learning in Risk and Investment Management

Tue, 05 Dec 2017 05:46:00 GMT           

In November 2017, Verne Global, together with the Professional Risk Managers International Association and The Realization Group, hosted a fascinating event covering Artificial Intelligence and Deep Learning in Risk and Investment Management, attended by over 70 industry professionals.

Going beyond the five V’s of Big Data
The event kicked off with Dr Oliver Maspfuhl of Group Credit Risk & Capital Management at Commerzbank AG. Dr. Maspfuhl discussed Big Data as “A Way Forward in Risk Management”. His view is that in order to use AI effectively in these types of contexts, firms need to capitalize on the “five V’s” that characterize big data: volume, velocity, variety, veracity, and value.

Volume and velocity refer to the staggering amount of data being generated by today’s businesses, and the rapid-fire rate at which data in areas such as payments arrives. Variety refers to the way data comes in many different shapes: It may be structured or unstructured, and unstructured data points or documents may be in very varied formats. Veracity refers to avoiding errors and problems in the data, which in turn enhances the data’s value for generating useful insights.

Today, Big Data needs to go beyond these base concepts. Big Data today is about understanding the causal linkages among data points. AI helps tremendously with that. It can visualize and analyze data in a way that people cannot do on their own. Given that, the role of the Risk Manager in the 21st century will very likely be to use artificial intelligence to identify new patterns, emerging risks, and new situations, and then to interpret them for maximum Value.

Robo Advising and Digital Wealth Management
Next, Mr Holger Boschke, Chairman of the Supervisory Board of TME AG, discussed “The Evolving Role of AI and Machine Learning in Investment & Wealth Management”. Mr. Boshke focused his talk on the fact that for financial firms, the computing power needed for AI is a back-end consideration; what truly matters is having an effective front-end product, and one example of this is robo advising. And robo advising is starting to evolve with emerging needs in digital wealth management.

Robo advising has grown very rapidly in the US. Since inception in 2013, robo advising grew to manage more than $8bn by 2016. As a percentage of the entire market either in Germany or abroad, robo advising still has a tiny market share, so there is plenty of room for the product to grow as AI improves. Indeed, robo advising is potentially a huge threat to the $1.3 trillion wealth management industry, because robo fees are much lower than traditional asset management fees.

The robo advising industry in some respects is paralleling the early days of ETFs, and there are similarities between the two products in that both save considerable amounts of fees for users versus traditional alternatives. To really capitalize on the benefits of robo advisors, though, banks need to overhaul their digital wealth management platforms by doing three things: (1) providing high quality relationship management; (2) offering the best tools in portfolio management; and (3) enabling customers to communicate with the bank easily. AI brings all of this together, from effective data-driven portfolio decision-making to enabling real-time speech-based digital communication.

AI computing infrastructure
Finally, Stef Weegels, Global Sales Director, Financial Services and Capital Markets at Verne Global, discussed “Infrastructure Considerations Around AI & Deep Learning Deployment”. Mr Weegels pointed out that many quant managers are turning to AI and deep learning for their trading, and they need ways to efficiently manage the data and computing that underpins this. Firms often need to go beyond the public cloud for their computing needs and look to data centers which can offer them dedicated private cloud infrastructure.
One of the key considerations around any AI computing infrastructure is the amount of power it consumes. In Iceland, where Verne Global’s data center is located, the cost of power is 80% cheaper than in Germany, and all of the power production is environmentally friendly and green. This means that Verne can offer clients pricing locked in for fifteen years at extremely attractive rates – an advantage that has led a number of leading financial firms to host computing for AI-driven alpha generation, risk management, and portfolio analytics with Verne.

Putting it all together
The event then turned to a discussion panel chaired by Mike O’Hara of The Realization Group. The wide-ranging panel discussion included questions from the audience on a variety of areas, including: the use of AI and ML in identifying different types of risk beyond market risk and credit risk; implications of upcoming regulations such as GDPR; feedback loops in unsupervised learning systems; identifying and adapting to regime changes in pattern-matching systems; use of deep neural networks in systematic and quantitative trading; AI in market sentiment analysis; infrastructure considerations when scaling out large AI computing platforms; and collaboration between banks and Fintechs in the AI/ML space.

In summary, this was a stimulating and idea-rich event, with some very thought-provoking discussions that continued well into the evening during the networking session afterwards.