San Francisco's AI Labs trumps Robo-Advice with neural networks

There's a lot of noise about how Robo-Advice is going to disrupt the wealth management sector. Indeed, the UK's FCA received over 40 applications last year to use automated wealth solutions of one sort or another. Many Robo-Advisors tout some level of machine intelligence to foster results, but really offer a generic, pre-set portfolio of three to six ETFs to clients.

San Francisco-based AI Labs began building its Vise platform in January 2015. Today it uses a mix of machine learning and deep learning.

Runik Mehrotra, president of AI Labs said: "To build portfolios, we implement a multitude of AI technologies coupled with conventional Markowitz finance. We believe there are underlying structures and patterns within finance and a combination of nonlinear and linear technologies can result in a more accurate system.

Dr Marc Ettlinger, chief data scientist, AI Labs, added: "We use innovative neural network models to build our sentiment analysis engine, far surpassing previous approaches for understanding the meaning of millions of stock articles and their relationship to stock price."

A huge innovation in data science over the past five years has been the ascendance of neural network models, rebranded as deep learning models, over symbolic, rule-based expert systems. There's a lot of hype and headline around this stuff just now: DeepMind beating Lee Sedol at Go, as well as the use of neural networks to solve important fundamental AI tasks such as image recognition, which is fueling innovations with self-driving cars and medical diagnostics.

Within a financial context, we hear about hedge funds hiring banks of PhDs with all manner of science backgrounds to search out alpha returns. However, many of these hedge funds (including some AI-driven funds) have performed poorly of late.

The Barclay Hedge Fund index performance can be classified as poor at best and downright horrible at worst. According to a hedge fund tracker HRF, net flows have been negative since the last quarter of 2015.

The problems partially result from a change in market outlook. With stocks at all-time highs and bonds at record low yields, investors in hedge funds are naturally looking for alternatives. Preqin's survey showed that hedge fund managers themselves acknowledge that there are fewer opportunities in credit and equity strategies. All the blame cannot be placed on the trading strategy itself but rather the market performance.

Another aspect of why hedge funds have been seeing poor returns, specifically those who tout algorithmic trading strategies and systematic trading, is due to their inability to continuously refine and upgrade their strategies, noted Mehrotra.

"Technology is constantly changing," said Mehrotra. "I was in the research sphere before I was at Artificial Intelligence Labs and I experienced how quickly technological research [quantitative finance, artificial intelligence, etc] was done and the effects and implementations of that research in systematic trading strategies.

"I still think large research platforms have a role in hedge funds. That role however, will slowly start to change as we see AI develop further. Their alternative strategy is really to look beyond stocks and bonds, increase the reputation of active investing, and strengthening their back office infrastructure," he said.

High commission hedge funds that capture and clean acres of unstructured data have been forced to face the fact that Robo-Advice and smart beta solutions have lately been achieving similar or even better returns. Unsurprisingly, this often elicits disdain from the active management and data science community.

At the recent Newsweek AI in capital markets event, Michael Beal, founder of Data Capital Management said that searching out alpha was fine, "but if you are packaging beta, it's about to be over for you".

Mehrotra acknowledged the recent traction around Robo Advisory tools and personal wealth management. Robo-Advisors such as Betterment and Wealthfront, for instance, have over $10bn in assets under management combined and are slowly catching up to some of the larger players in the industry such as Vanguard and Schwab.

He said: "They've been yielding market-like returns for quite some time now. It's interesting to look at their correlation with the market and how exposed these portfolios are to the indices, though it's far more interesting to look at their returns in market downturns."

During Brexit, for example, Betterment portfolios fell an average of 3-4%. Consequently, Betterment halted all trading for the day. For a product that markets itself as having control over your portfolio, that move was very hedge fund-like, pointed out Mehrotra. Smart beta or Robo-Advisory practices might be beneficial in the short term, but in the long term, especially during market downturns and periods of high volatility, it will be difficult for Robo-Advisors to justify their fee.

Samir Vasavada, CEO, AI Labs added: "So far, Robo-Advisors have emphasised their fee-savings, while neglecting the long-term attractiveness of the asset allocations they propose. Instead they argue for a simple passive approach to investing that works for seven years and fails miserably for two or three.

"They've been fortunate that the market has benefited them for those past seven years - but the shortfall of their business models will only be felt in the next market."