Alpesh Patel, CEO of private equity firm Praefinium, recently conducted two experiments to see if OpenAI's GPT-4 language model could be useful in picking stocks and potentially improve the returns of his personal portfolio.
What Happened: Patel experimented with GPT-4, reported Insider. His investment portfolio consisted of around 20 stocks. He wanted to find out if the stocks he picked would match the choices of GPT-4.
Patel fed GPT-4 65 data points from a terminal. These data points included working capital, free-cash conversion, debt, pre-tax profit, normalized EBITDA, capital expenditure, cash return on capital invested, stock volatility, and even his firm’s stock ratings that weigh valuations, growth, and income.
However, after these experiments, Patel concluded that while artificial intelligence, or AI can assist in removing some human bias from the stock-picking process, it does not have as good a track record of generating greater results.
Experiment 1: Patel used historical data from December 2020 to December 2021. He prompted GPT-4 to predict the five top-performing stocks for the next 12 months, up until December 2022.
The results showed that GPT-4 was unable to pick stocks that would have performed well in 2022, and the top three it selected — Bouvet ASA, Jarvis Securities, and Cake Box Holdings — didn't trade on U.S. exchanges and even suffered substantial losses beyond 20% in 2022.
Experiment 2: Here, Patel experimented in real time. He fed GPT-4 data points for the 30 Dow stocks in the 12 months through May 3. He prompted it to select the top five stocks it thought would perform well over the ensuing 12 months.
GPT-4 picked three of the same stocks Patel already held — Microsoft, Visa, and Apple, all of which had significant returns.
Why It Matters: The launch of chatGPT for public testing last year sparked a lot of buzz around the generative AI technology. But Patel's experiments show that although AI can help with removing some human bias from the stock-picking process, its scorecard for producing superior returns is not as strong. He's also concerned that attention-grabbing AI tools might be employed by marketers to mislead investors.
Patel says that AI doesn't "tell us the future as it were," but sort of confirms "what we almost know and should pay more heat to," which is the need to be a "bit more humble" and "not so overconfident" in front of the data.
However, he adds that such experiments could also serve as valuable tools for educating people on how to be better investors by understanding how to use AI and the importance of having a robust approach to stock picking.
He says ChatGPT will not result in job losses. On the contrary, he feels "it's going to lead to people being retained because now I’ve got a bunch of people who I never thought about educating on finance to this level that I can."
Read Next: Student Accused Of Plagiarism With ChatGPT Turns To Reddit For Help: Here’s The Advice They Got
© 2024 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
Trade confidently with insights and alerts from analyst ratings, free reports and breaking news that affects the stocks you care about.
