The Cambridge Handbook of Artificial Intelligence (AI) defines AI as a cross-disciplinary approach to understanding, modelling and creating intelligence of various forms. According to Konstantine Arkoudas and Selmer Bringsjord, it is a field devoted to building machines capable of displaying behaviours deemed intelligent, at least in well-controlled environments.
A machine that possess intelligence similar with or superior to that of a human poses numerous ethical and legal issues. For example, Nick Bostrom in Superintelligence: Paths, Dangers, Strategies wonders how AI would see the human values and purpose. He suggests that the idea of a machine (with its essence being algorithms) is incompatible with the biological nature of our feelings that set the base for our moral values. A similar ethical consideration is posed by Matthias Scheutz in Artificial emotions and machine consciousness. The author explores the degree to which a machine is capable of not only feelings but the ability to make a distinction between wrong and right in situations that lay outside its given parameters.
The above subjects are worthy of entire books to be written on them. However, for the purposes of the article, we will not pursue philosophical questions. Instead, we are concerned with one particular dimension of AI – learning. More specifically, we are concerned with machine learning and its impact and use within the financial sector. Machine learning is defined by David Danks as;
“A set of algorithms and techniques that have been applied to problems in a wide range of domains.” tweet
Therefore, machine learning is a subfield of computer science that draws on models and methods from statistics, algorithms, computational complexity, control theory and other disciplines ‘to infer predictive models from large data sets’, as suggested by the authors in Machine Learning for Market Microstructure and High Frequency Trading. Consequently, it is their ability to navigate copious amounts of data and identify patterns that make the AI useful to finance (at least for now).
High Frequency Trading
High Frequency Trading (HFT) is defined by Tom Lin in The New Financial Industry (March 2014) as a type of algorithmic trading characterised by high speeds, turnover rates and order-to-trade ratios that employs large amounts of financial data and electronic trading tools.
HFT is used by traders to profit by trading large amounts of assets in a very short period of time. Also, HFT has a number of features which we will describe here that, at least for now, might seem to give you ‘the edge’ in beating the market. For example, according to Yuri Nevmyvaka, Yi Feng and Michael Kearns (Reinforcement Learning for Optimized Trade Execution, 2006) you might be able to optimise trade execution via reinforcement learning (a method which will attempt to specify the optimal action to take from a given state) and thus you might receive recommendations whether to be more aggressive or defensive with your market orders. This can reduce trading costs tremendously.
Moreover, AI can be used to predict price movement from market patterns (cycles). This strategy is not aiming to save trading costs by deciding when to execute a given trade as the above strategy but rather it will provide you with signals of when to trade and how to trade.
However, this seems too good to be true, does it not? Indeed it does. Remember the Flash Crash of 2010? It was the result of HFT. The situation saw the US trillion-dollar markets collapsing and rebounding extremely fast. It was caused by a large number of trades placed and executed rapidly (in nanoseconds or milliseconds) that concluded into market volatility close to an anomaly. At the core were a class of financial instruments engineered using digital technology – the Exchange Traded Funds. Without getting into the details of the crisis, HFT (the AI that helped traders cash in billions) exposed to permanent capital risk thousands if not millions of investors as the prices of stocks, futures, options and ETFs dropped to a penny per share.
The Flash Crash posed a crucial question as to: Whois legally liable? Some argued that the traders were not guilty of misconduct as they have followed the recommendations of an algorithm. However, it was concluded that even if they have followed the recommendations of a machine, they were rationally capable to decide to act on such recommendations or not. Also, the regulatory framework is ancient and deeply flawed and unequipped to deal with AI.
AI and Finance: The bigger picture
AI does not need to directly shape our financial industry. According to an article by FT, published on January 4, 2015, AI attracts a lot of money from investors. For example, Kensho, an AI company, raised USD 15 million in order to train computers to replace expensive white-collar workers such as financial analysts. Moreover, Sentient Technologies raised USD 143.8 million to pursue its data analysis through massively scaled AI and Nara Logics raised USD 13 million for its big data analysis programme.
Other firms, such as Charles Schwab, took it even a step further in using AI and launched a new service called Schwab Intelligent Portfolios. Obviously, it uses machine learning to build tailored portfolios and remove management fees.
If one were to allow their imagination to run wild, one can imagine how AI can improve accounting principles: imagine never having to match receivables, net income and cash flows in order to uncover fraudulent management. A machine will have no incentive to cheat GAAP or other accounting standards, it will simply comply with them. Also, perhaps a world where human greed and envy play no role in economic decisions such as resource allocation because numbers from an algorithm will do so. Of course, cyber security risks need to be considered. But this is a topic for another time to be explored.