It’s no secret that AI software such as Open AI’s Chat-GPT, Google’s Bard, and Microsoft’s Bing AI have created immense opportunities for individuals. Naturally, people have attempted to use these technologies for profit. This leads to the question, how good are these programs at selecting stocks or strategies that will outperform the market?
Sample Tests:
To test this, a UK based firm, Finder.com, asked Chat-GPT to compile a list of stocks it thought would outperform the market. The list included stocks ranging from behemoths such as Amazon to lesser known, but still massive, companies such as Lam Research (a $92 Billion Market Cap). In a period between March 6th and April 28th, the fund rose a respectable 4.9%. However, the average of 10 leading investment funds actually reported a loss of 0.8%. This must be a fluke, right? No, another study asked 2 different AIs (Google and Microsoft) to select 2 stocks to invest in. Google’s Bard selected Microsoft (surprisingly!) and Walmart, Bing selected Verizon and Shopify, and, for a control, the researcher picked Lockheed Martin and Campbell’s soup. Over 3 weeks, both AI’s outperformed the S&P 500 but only Bard was able to beat the reporter.
What to make of it:
The data looks promising for the machines, but let’s make sense of it. These studies were only conducted over a few weeks, but that is not enough time to make a conclusive decision. These large indexes include many stocks and are not focused on the month to month performance of a company because it is a major decision to remove a company from the S&P 500 based on just that data. Therefore, even if a company is not doing so well, the fund will still include it, diluting its performance amongst the other companies they keep. AI, on the other hand, is more likely to pick big-name, large cap companies which have a smaller chance of weak performance. The probability of randomly selecting two stocks that will outperform the average is high, especially with the AI’s bais. There is a Youtube video in which a goldfish is given $50,000 to trade stocks depending on where in the fish tank they are in. Unsurprisingly, given the context, the fish outperformed a sentiment based model that could choose any stock, as opposed to the “safer” stocks the fish could choose from. Clearly, the way that we analyze these results needs to be considered.
So what now:
It may seem that AI guides us to a treasure that we should not possess. But that is not the case. There are still plenty of safe ways to use AI to your stock-picking advantage. You can ask for it to define financial topics and terms in a simple way. For example, try asking to to define high-frequency trading as if it was speaking to a 5 year old. It can also help you find strategies or indicators that match your investing style and you can also ask what strategies would work for your situation. Think of it as a highly advanced Google rather than a machine that knows it all - the way that you use the information is what matters.
Sources
