Artificial intelligence (AI) is increasingly used in many areas, including language translation, medical diagnosis, and so on. Of interest here are AI applications in stock market activities. Most investors are aware of programmed trading and high-frequency trading (HFT), both of which use computer algorithms to decide on buying and selling equities. One can expect the further expansion of AI into equity markets, as described in this book. The mention of “Skynet” in the title may carry strong negative connotations for those familiar with the Terminator movies, but the overall view of the authors is quite positive.

We find here a high-level overview of economic theories, including the philosophical underpinnings, from Adam Smith to today, combined with the authors’ view of how AI will impact and change market performance. The authors begin with supply and demand, rational choice, and Simon’s theory of bounded rationality, which posits that economic choices can only be partly made using rational thought due to the limited availability of information. Behavioral economics considers how human nature, emotions, and judgments based on incomplete data affect decisions. For example, economic theory assumes that individuals participate in markets to gain “utility,” that is, to improve their situation. Information asymmetry--the idea that parties to a transaction share differing data about the object being transacted--reduces market efficiency because the party with better information gains an advantage over the other party. Game theory introduces a system of rules used by multiple players seeking maximum utility. The rules can be used to build models of transaction outcomes, permitting economists to test their theories. Pricing and value theory, briefly introduced earlier in the book, is treated here with regard to how different AI methods improve the effectiveness of valuation and market efficiency. Mechanism design--in a sense, a reversal of the game theory model--creates what-if scenarios that work back from the desired outcome, maximum utility, to define the rules that result in that outcome. The efficient market hypothesis states that an individual cannot outperform the market as a whole because the market possesses and uses all available information, beyond an individual’s capabilities. Portfolio theory seeks to optimize an individual’s tolerable level of risk to achieve a certain return on investment (ROI). Financial engineering applies engineering principles to financial problems. The classic example of this is the Black–Scholes pricing formula for European options. A set of appendices provides brief descriptions of the AI and optimization techniques discussed in the text.

How does AI affect these various elements of economic theory and behavior? The authors’ view is that AI systems can use big data and/or optimization techniques to achieve better rational analysis and decisions than individuals. They can reduce information asymmetry and enable more individualized pricing, thus improving market efficiency. Machine learning combined with optimization techniques, such as particle swarms, simulated annealing, and genetic algorithms, permits AI systems to optimize their processes and adapt to changing market conditions. This all seems quite positive. Are there any downsides? For example, could rational AI systems operating in a universe of irrational human decisions lead to lowered efficiency, such as the flash crashes that have already been observed when panic sales by people lead to parallel algorithmic sales? How would the use of different AI systems with varying capabilities affect information asymmetry? How would markets be affected by an AI system (that is in current use) that finds obscure trading patterns that yield profits for the owner but in so doing reduce market efficiency? If AI systems approach levels of human intelligence, might unpredictable emotions and irrational behavior emerge? The authors should at least have shown that they considered these possibilities, if only to show why they can be dismissed.

The authors present a clear, high-level overview of basic economic theories interwoven with relevant AI applications. They provide copious references for those readers interested in details. The simulation results that drive the conclusions are somewhat lacking in detail. More direct references to sources would be helpful, particularly for the charts containing simulation results. It is sometimes difficult to tell if a chart is a generic illustration or presenting specific data. The authors show how AI systems have improved, or could improve, market efficiency and utility for participants. Although the material is interesting and thought-provoking, I would have preferred a discussion of the possible negatives, such as those noted above, the Skynet aspects of AI in the markets.

More reviews about this item: Amazon