Tools like ChatGPT, with artificial intelligence (AI) built right in, have the potential to revolutionize how quickly, efficiently and effectively people complete their work. And this holds true for all areas of our lives, including industries like healthcare, manufacturing and most others. The financial markets are no exception.
Although AI has many advantages, there are also potential risks, as the expanding usage of these technologies in the financial sector demonstrates. Important lessons about the repercussions of employing computers and AI for decision-making can be learned from an examination of Wall Street's earlier attempts to speed up trading by embracing such technological advances.
Early in the 1980s, institutional investors started employing computer programs to carry out transactions under predetermined rules and algorithms, spurred on by technological breakthroughs and financial innovations like derivatives.
This type of “program trading” became more complex as technology improved and more data became available, with algorithms able to analyze complex market information and execute trades based on a variety of criteria.
Black Monday
This ultimately led to the Black Monday stock market crash of 1987, during which the Dow Jones Industrial Average (DJIA) experienced the most significant percentage collapse in its history. The anguish was felt all around the world and reciprocated for some time.
Fast-forward 15 years, and in 2002, the New York Stock Exchange unveiled a completely automated trading platform. Program traders were replaced by “high-frequency trading” (HFT), which used significantly more complex automation and cutting-edge technology.
HFT analyzes market data and quickly executes trades using computer programs. High-frequency traders use powerful computers and high-speed networks to analyze market data and execute trades at breakneck speeds. This is in contrast to program traders who bought and sold baskets of securities over time to profit from an arbitrage opportunity — a difference in the price of similar securities that can be exploited for profit — by buying and selling baskets of securities. Compared to the several seconds it took traders in the 1980s, high-frequency traders can now execute trades in around 64 millionths of a second.
These transactions may include purchasing and selling the same security several times in a matter of nanoseconds and are typically of a very brief duration. Artificial intelligence systems analyze vast volumes of data in real-time and spot patterns and trends that human traders may not be able to discern right away. This aids traders in making wise selections and carrying out trades more quickly than would be feasible manually.
Natural language processing, which entails analyzing and understanding human-language data such as news articles and social media posts, is another significant application of AI in HFT. Traders can improve their understanding of market moods and make necessary adjustments to their trading strategies by analyzing this data. These high-frequency, AI-based traders behave substantially differently than their counterparts from yesteryear.
The Negative Side of a Powerful Combination
Efficiency and speed, while seemingly beneficial to the industry, can both be harmful in their lack of prudence. HFT algorithms are capable of responding to news events and other market signals so quickly that they can create abrupt increases or decreases in asset prices.
Furthermore, HFT financial companies may exploit their speed and technology to outperform other traders, thus skewing market signals. The so-called “flash crash” in May 2010 happened because of the volatility caused by these AI-powered trading bohemoths. Stocks dropped and then recovered in a matter of minutes, wiping out and then recovering around US$1 billion.
Due to how quickly and effectively high-frequency traders analyze the data, even a slight change in the market's dynamics can result in a huge number of trades, which can cause sharp price fluctuations and increased volatility.
This introduces us to a new universe of trading algorithms and related software driven by ChatGPT. These can exacerbate the issue of having too many traders on one side of a contract.
When left to their own devices, humans typically make a wide variety of choices. However, such rapid and all-available technology may reduce the range of viewpoints if everyone is using the same artificial intelligence at the same time to draw their conclusions.
Decisions made by generative AI-powered chatbots would be similar to one another because they are based on prior training data. ChatGPT is very likely to recommend the same brand and model to everyone. This could intensify the problem of herding and result in substantial price increases as well as shortages of some goods and services.
Furthermore, when the AI making the judgments is given inaccurate or biased information, this becomes troublesome at its core, as such systems trained on skewed, outdated or insufficient data sets have the potential to reinforce preexisting prejudices. Working from this, ChatGPT and similar technologies have come under fire for factual mistakes.
At least for the time being, most US banks appear reluctant to permit their staff to utilize ChatGPT and other tools. Once they address their reservations, however, they may finally accept generative AI and the risks it brings.
