Are regulators ready for robots?

In October 2015, two of the world’s largest companies, Google and BlackRock, announced plans for a joint venture. The purpose of this formidable partnership is simple — to explore how artificial intelligence (AI) can be used to improve investment decision-making.

The collaboration is the latest in a series in which major financial institutions look to harness new innovations and techniques from across a range of inter-related disciplines including; cognitive computing, machine learning, deep learning, neural networks and AI. It has led to fierce competition over the best data scientists, with high-profile names like Bill MacCartney moving from Google to BlackRock, Alfred Spector moving from Google to the hedge fund Two-Sigma and the former chief engineer behind IBM’s supercomputer, Watson, joining Bridgewater Associates.

The deployment of these brilliant minds to the task of investing, will most likely lead to a wave of new innovations. It may also make some organisations unbelievably profitable. But it will also continue to make financial markets even more complex and arcane — further restricting knowledge of its inner-workings to an ever-shrinking group of technical experts.

This is nothing new. Technologies like AI and machine learning are just the latest in a decades-long process, which has seen the wholesale transformation of financial markets. What were once physical marketplaces, predominantly driven by instinct, intuition, rumour and emotion, are now digital markets, populated by black-boxes and fed on a diet of massive, unstructured data.

One question which arises is whether financial regulators are able to maintain an informed understanding of the markets, in this brave, new, black-boxed world.

The popular sentiment suggests that they are not. In our recent study exploring the future of Artificial Intelligence in financial markets, out of a global survey of financial executives, 69% of respondents are either “not very confident” or “not confident at all” in the regulators understanding of financial technologies. Seventy-six per cent either agreed or strongly agreed with the statement, “Financial regulators’ knowledge is not keeping pace with advances in technology”.

Is this lack of confidence in regulators fair? Regulators are undoubtedly behind the private sector in their comprehension of data science. However this is largely due to a huge disparity in resources. Consider that in 2014/2015, the annual budget of the UK’s main financial regulator, the Financial Conduct Authority, was £452 million. In 2012 Google spent nearly the same amount (£400 million) acquiring the company Deepmind Technologies — a firm which employees over a dozen of the world’s leading experts in machine learning. BlackRock also spent £200 million, acquiring the robo-adviser firm, FutureAdvisor. Clearly, financial regulators are outgunned in human, financial and technological terms.

Why should this be a problem? It’s not the private sectors fault that it has accrued astonishing profits, granting them superiority in technical know-how and equipment. However there is a danger that lessons learned from the most recent (and ongoing) financial crisis, will be forgotten. The crisis that culminated in 2008 demonstrated how “light-touch” regulatory authorities, equipped with inadequate statutory powers and an insufficient understanding of the financial system, were in no position to recognize the build-up of systemic risks. As former Chairman of the Federal Reserve, Ben Bernanke, conceded in a lecture in 2010, “it was difficult for any single regulator to reliably see the whole picture of the activities and risks of large, complex banking institutions.”

Whilst still only fledgling, the introduction of sophisticated AI trading programmes, employing advanced techniques such as machine learning, presents regulators with components they are not in a position to understand. Indeed even within leading systematic investment firms, only a handful of individuals will have an understanding of what really goes on inside complex trading algorithms. It seems fair to conclude that, in the event of any programme malfunctions, regulators will be swiftly out of their depth.

How should regulators make-up for this disparity in technical skills and knowledge? The survey respondents identified two measures regulators could pursue. The first is to encourage more collaboration between regulators and fintech companies (32%). The second, is to coordinate regulatory efforts across markets, in a systemic global fashion. (25%)

These are useful initiatives, which are already being explored by a number of financial regulators. For example, the Financial Conduct Authority (FCA) in the UK has launched Project Innovate, which encourages FinTech companies of all shapes and sizes, to collaborate with the FCA in the development of new products.

But collaborative relationships can soon turn into cozy relationships. This is why regulators might need to explore other initiatives, such as surveillance of a private firms’ technology — in particular, the source-code of a firms trading algorithm. This initiative is being actively explored by the US Commodity Futures Trading Commission, under their proposals for the Regulation of Automated Trading (sometimes known as ‘REGAT’). If approved, this will give US regulators unprecedented access to a financial firms’ trading algorithm.

The question of whether or not this initiative oversteps the mark, is for another discussion. However what this initiative exposes is the tension that exists between the ineluctable desire of the private sector to keep pushing for growth and innovation — sometimes with little consideration for the wider costs, against the restraining responsibility of regulators and governments to ensure that innovation is stable and safe.

The irony is that, whilst new technologies pose great risks and uncertainties for regulators, they also have the potential to empower regulators to an unprecedented extent. AI and machine learning enables small organisations to gain a deep understanding of highly complex systems, by crunching lots of data. This is precisely the technology that could help regulators better understand the complexity of the markets and possibly even anticipate the next crisis.

Here’s hoping that Google’s next joint venture is with the regulator.