How can AI and ML drive hedge funds?

Investing in AI

While hedge funds have long utilised computer programmes to make trades - quants have been utilising statistical models to find market patterns as part of their strategies for decades - hedge fund investing is evolving to include artificial intelligence (AI), technology that attempts to emulate human behaviour, in much more of the process.

In the case of some hedge funds, AI has taken the reigns and with more big funds embracing the technology, there is an indication that it may be gradually moving into the mainstream.

Not to be confused with high-frequency trading (HFT), which uses powerful computers and algorithms to trade potentially huge numbers of orders at incredibly high speeds, AI is increasingly being adopted by hedge funds to generate investment ideas, help with portfolio and investment decisions, for signal generation, to manage risk, and in some cases execute trades.

The use of the technology in investment circles is not an entirely new concept and some managers have had their eyes on AI for hedge funds for years. An early adopter of machine learning (ML) in hedge funds, Jim Simons, one of the world’s most famous hedge fund managers and founder of Renaissance Technologies, has been using similar models to help formulate his investment strategy for decades.

While hedge funds that exclusively use AI are still relatively rare, many are using AI or ML in at least some element of their business. As interest from managers grows, and as the technologies underpinning these strategies continue to develop, what might AI mean for how hedge funds invest?

Artificial intelligence investing

In a 2018 survey by BarclayHedge, 56% of respondents said they used an AI/ML to help inform their investment decisions, while around two thirds indicated that they looked to the technology to help with investment ideas and to optimise portfolios. A quarter of the respondents said they used AI or ML to execute trades.

A factor behind the increasing interest and use of AI and ML in hedge fund investing is the ever-increasing growth of data. BNY Mellon points to the fact that more data is now generated in a day than in the entirety of 1990.

Big data’s proliferation has lowered costs for the capture of data, making the datasets needed for complex AI and ML processes more available and at lower prices. Meanwhile, computationally demanding ML processes have benefited from the reduced cost and growing availability of computational power , and the growth of the cloud has made available more processing power.

Machine learning in hedge funds

A subset of AI, ML learns to find and develops an understanding of patterns in data with no specific instruction, or without being told what to look for .

ML can be used by hedge funds to identify nonlinear relationships between assets, to find new opportunities in uncorrelated assets, and to potentially uncover entirely new relationships , presenting new and unforeseen diversification benefits.

ML programmes can process huge amounts of financial data, historical and real-time, and map it in a way that the human brain is unable to , potentially unveiling new investment angles or strategies.

AI and ML proponents suggest that for hedge funds to stay competitive and continue to find small dislocations and inefficiencies in the markets will require more and more advanced analysis of market data.

Although it's notoriously difficult to track the performance of funds utilising AI for investing, the Conversation points to academic research that has found examples of highly accurate financial forecasts based on ML models .

ML programmes’ capacity to process huge amounts of data at incredible speeds vastly increases the volume of data that can be analysed by a fund and reduces time and cost constraints associated with human resourcing, as well as potentially reducing human subjectivity errors.

As well as providing advanced analysis of fundamental financial data, ML programmes are being applied to analyse various alternative data such as transactional data from credit cards and point of sale, Internet of Things records, and satellite imagery.

Further, natural language processing, a form of ML, can be used to analyse unstructured data such as from news and social media feeds, and earnings calls to gauge sentiment , as well as to reveal other potentially advantageous insights .

Event proofing hedge fund investing

Gaining increased insight from alternative data can help hedge funds gain an edge on competitors, and proponents of AI and ML point to the advantages of these processes in analysing and finding patterns and looking for signals in this massive pool of data.

Some proponents even suggest that advanced AI models will not even need the expert knowledge of financial markets that hedge fund managers have developed over decades, as the programme will itself be able to autonomously develop from the data “nonlinear statistical relationships undetectable to human-based and traditional ML methods”.

There is an argument that the wealth of data available, and the advances in the AI programmes used by hedge funds, have made investing models using the technology better at adapting to unexpected global events, such as a global pandemic , than traditional quant models and traditional actively managed funds.

In addition to helping hedge funds with their direct investment objectives, AI can also be used to assist and improve efficiencies in other supportive processes . From hiring to compliance and other back office functions , utilising AI could help improve efficiency and lower operating costs, which in a low-fee environment is an enticing proposition for funds.

The limitations of artificial intelligence investing

Whereas AI and ML processes have excelled at learning how to play Chess or Go, or in helping to train autonomous vehicles, financial markets and the information used to build investment strategies present a less structured set of rules on which these programs can learn. These programmes also need a huge amount of data to be properly learn , which financial markets may not be able to provide.

Furthermore, claims of ML programmes' ability to map patterns in financial markets should be considered carefully, as patterns recognised by intelligent machines can be too broad , i.e. finding correlations that don’t necessarily have a real-world investment application, or the patterns discovered may already be known and recognised, thus presenting a wasted effort.

For managers looking to invest in artificial intelligence, caution should be taken in understanding the current limits. At least for now, AI and ML may not be a shortcut to fantastic returns but can be incredibly powerful tools for improving investment decisions nonetheless.