AI Daily - Artificial Intelligence News

View Original

How Covid-19 exposed algorithmic faults

One of the major impacts that Covid-19 has in our society was definitely the stock market. It may seem unchanging or consistent nowadays, but back in March, the true and horrible brunt of the virus was felt. The month of March 2020 was apparently the craziest in stock market history, beating out crashes in 1929, 1987 and even 2008! However, March 2020 is not necessarily the worst performing month in stock market history, it ranks 16th on that list; there was only a 13.7% decline in the Dow Jones Industrial Average, the benchmark that we would use to measure financial disparities.

Therefore, due to its volatility and relative unpredictability, March exposed some of the weaknesses in quantitative trading firms (Also known as quant firms), wherein they would rely on mathematical models, including ones which implemented artificial intelligence, to make trading decisions for them. It was catastrophic. Many quant firms experienced a loss, such as Bridgewater Associates – their funds dropped almost 21%, according to a statement posted by their co-chairman on LinkedIn (link it). This was shown to be inevitable, as Ray Dalio, their chairman spoke of their tracking the virus since January, but the common factor between any of these quant firms is for example in the case of Renaissance Technologies, they admitted to their investors that the algorithms they used misfired in response to the volatility of the month.

However, March 2020 is not a complete anomaly; Professor Andrew Lo, MIT, recounts that March bears similarities to a meltdown among quantitative firms in 2007, in the early days of the financial crisis. “What we saw in March of 2020 is not unlike what happened in 2007, except it was faster, it was deeper, and it was much more widespread,” Lo says.

Credits: Shutterstock

We can attribute the aforementioned turbulence to a possible limit in the capabilities of modern-day AI, which in itself is built around targeting and exploiting subtle patterns in swathes of data, something that would take exponentially longer with a human brain. Just as the algorithms that were employed by grocers to stack shelves were perplexed by the consuming costumers’ seemingly random obsession with purchasing toilet paper, the algorithms that help the hedge fund companies wring profits from the market were equally confused by panicked investors and their accompanied instability. Essentially, one might remark that the algorithm is only as good as the data it’s fed, and since we had no way to predict people’s spending and reaction to news and advice in this Covid-19 period, there was no way our algorithms could have either. We can therefore assume that artificial intelligence will not be independent of humans for a while, as this pandemic has proved. And we can expect investors to cast a wider net for data to feed to their algorithms, in hopes of detecting signals of unusual economic activity.

Thumbnail Credit: pintrest.com