How AI Can Improve Our Investment Capabilities

Many investors today believe that artificial intelligence will be the investment panacea, but this is not the case.

‍

Don't just believe what I say. Just ask ChatGPT, the OpenAI AI program that has recently generated so much buzz. It replied that this "is a hard goal" and that "previous success is not a guarantee of future results" when I asked it how to beat the stock market.

‍

Of sure, ChatGPT is correct. But compared to the exaggerated expectations many investors have for what AI can accomplish in the stock market, its prudent counsel falls far short.

‍

The Eurekahedge AI Hedge Fund Index's performance, which "is designed to provide a broad measure of the performance of underlying hedge fund managers who utilize artificial intelligence and machine learning theory in their trading processes," is another reason to be skeptical. Since 2009, this index has lagged behind the S&P 500 in cumulative performance.

‍

Of course, comparing the success of these AI-informed hedge funds to the S&P 500 may not be the best comparison. For instance, the Eurekahedge index performs substantially better when risk is taken into account. However, the majority of the investors who are so enthusiastic about AI's supposed ability to outperform the stock market are more interested in the immediate profits than the more ethereal benefits of the Capital Assets Pricing Model.

The increasing evidence that stocks related to AI are experiencing a craze is a third reason to be suspicious. Think about a study that Chih-Chiang Wu of Taiwan's Yuan Ze University and Wei-Peng Chen of the National Taipei University of Technology just published in Finance Research Letters. They discovered that everything else being equal, an ETF with "AI" in its name would perform better if it owned a stock than one without. This reminds me of the dot com bubble when companies' stocks would perform noticeably better when their names were modified to contain the "dot com."

‍

However, we shouldn't discount a strategy simply because some investors overestimate its potential. I contacted Doron Avramov, a finance professor at the Interdisciplinary Center (IDC) in Herzliya, Israel, who has authored numerous papers on AI, to acquire a deeper understanding of the technology's true potential. He stressed the following in an interview:

‍

In situations when the signal-to-noise ratio is exceedingly low, such as when predicting future returns on stocks, cryptocurrencies, or other financial assets, AI is not well suited. Another challenge is the remarkable efficiency of the markets, which makes it unlikely that any profitable new strategy developed by AI will last for very long. Therefore, even with an AI-discovered technique, there is no way to predict whether it would be effective until you start using it.

‍

• Investors who believe AI will be able to anticipate the direction of the economy or the stock market on a constant basis are deluded. The best chance for AI is to find two prices that are only slightly out of sync with one another and where this misalignment can be taken advantage of from a large collection of assets.

‍

• James Simons of Renaissance Technologies has used this arbitrage strategy to great success; between 1988 and 2018, his Medallion Fund had an average return of 39.2% (after fees), compared to the S&P 500's "only" 10.0%. Even though it appears incredibly promising, keep in mind that only slightly more than 50% of these arbitrage deals are profitable and that the typical holding period is frequently measured in seconds. You would need to carry out a massive number of trades in order to profit significantly from the strategy. Individual retail investors cannot hope to replicate this strategy.

‍

• A limited amount of money may be traded using the AI approach, and there are decreasing returns to scale from the Renaissance Technologies approach. The Medallion Fund is therefore solely accessible to the partners of the company and inaccessible to the rest of us.

‍

All of this does not imply that AI will never be able to increase our chances of outperforming the market. Professor Avramov is still looking at different AI-based tactics, some of which he claims are showing potential. But for the time being, it's critical to control our anticipation and excitement.

‍

ChatGPT is correct: Even with AI, outperforming the competition is still a difficult objective.