How people are using AI for stock market picks
Share Now on:
The popularity of ChatGPT has exploded since the artificial intelligence chatbot was released to the public last fall. In just a matter of months, it’s gained more than 100 million users.
It can write haikus, pass law school admissions tests and help plan your dinner, but can it make you money in the stock market?
It’s a prospect a lot of people are intrigued by, according to a new survey from The Motley Fool.
The investment advice platform polled 2,000 Americans about their interest in using ChatGPT for picking stocks.
Asit Sharma, a senior analyst with The Motley Fool, says the practice is already widespread. Marketplace’s Meghan McCarty Carino recently spoke with Sharma about the survey and his analysis of the results.
The following is an edited transcript of their conversation.
Asit Sharma: We found that nearly 50% of Americans [surveyed] have used ChatGPT for stock recommendations. We should say that we didn’t ask really specific questions about how people are using ChatGPT. But I can tell you from what we’ve observed from interacting with our members, people are using it in all sorts of ways. They’re asking ChatGPT to give, say, a list of small-cap stocks in the biotech space, or “Hey ChatGPT, tell me about a key risk that’s associated with investing in, say, Nvidia.” Or “Could you tell me some safe stocks to invest in in a choppy market?”
Meghan McCarty Carino: Now, was this a case of people who were new to stock trading using this as sort of a beginner’s guide or something else?
Sharma: So we’re seeing really interest across a broad spectrum of investors. We’re seeing it from people who understand that this could be a technology that provides you with a more robust response than just Googling a question about stocks. And we’re seeing very sophisticated investors who’ve been around the markets for a long time who are cutting and pasting, say, financial statements into ChatGPT and asking for very specific insights to see if there’s something that they might have missed in their own analysis.
McCarty Carino: You broke down some sort of demographic trends in this. Can you tell me about some of the trends that were noted in this survey?
Sharma: Sure. So I mentioned before, 47% of Americans have used ChatGPT for stock tips, that’s the nearly 50% I referred to. Here’s an interesting one: 77% of high-income Americans have used ChatGPT to try to suss out some stock recommendations. This makes sense to us because we’ve always observed over the years that higher-income Americans typically have access to tools before lower-wage earners do, and also they’ve got a little bit more leisure time to test out the latest technology.
McCarty Carino: Who in the survey was more likely to be engaging this technology and who was a little more skeptical about it?
Sharma: So we found that while 47% have used ChatGPT for stock picks, the number of lower-wage earners who have used it is still, I think, to be desired. Men are more comfortable using ChatGPT than women. Fifty-one percent of men are more comfortable relying solely on ChatGPT for investing recommendations versus 41% of female respondents who are more deliberative and less likely just to jump headlong into an investing trend. Millennials tend to be fairly trusting of ChatGPT, but they’re still skeptical, as are Gen X and Gen Z respondents. But baby boomers are the least likely to express confidence in the accuracy and trustworthiness of ChatGPT stock picks. And I’ll reflect [on] that. You know, baby boomers started investing in a day and age where you picked up a rotary phone to call your broker to place a trade with a high commission. They have seen every single evolution to the present day from those days. And they’re naturally skeptical of every new wave of technology, not that they’re superlate adopters, they just have seen it all. So they are the least likely to be confident in the accuracy and trustworthiness of these picks.
McCarty Carino: What are the risks in relying on tools like this for investment advice?
Sharma: One of the risks I think most people will be familiar with is the tendency of ChatGPT to hallucinate, that is to create answers that have no basis in reality. We know that ChatGPT is not the most accurate purveyor of results because it wasn’t built to be a search engine. It’s a large language model. So we should be careful about assuming that any result ChatGPT gives us is the gospel truth. We’ve got to be a little skeptical of what we’re reading. I think another risk is a risk that the average investor may take ChatGPT’s insights for the only insights or the correct insights. It’s just trying to pick out what might be the most probable outcome. ChatGPT might relate to you an investing insight that isn’t the most important one to focus on.
McCarty Carino: Yeah, I think the concept that this is a large language model but it’s sort of predictive text is an important thing to understand when you’re thinking about, you know, the data analysis that might go into investment advice, right?
Sharma: Yeah, totally. There’s something too that we should say could be very powerful in the future. The nature of ChatGPT is the transformer model. It places attention on what it sees is important in a certain context versus trying to understand sequentially what might be most important. This is really interesting when you think about investing because investors tend to look for patterns. Every investor has his or her own idea of what makes a great stock. The longer you play this game, the more you’ll have a personalized view of that. ChatGPT, large language models, are really, really good at picking out patterns. If you feed them what you’re looking for, they have this innate ability to see patterns to bring together disparate qualities and say, “Hey, Meghan, take a look at this company because this might fit what you’re looking for.”
McCarty Carino: What [could] tools like these miss that a human analyst like yourself might not?
Sharma: So these models are really great at scraping information from different sources. They’re good at sentiment analysis, so they can pick out different emotional indicators in earnings call transcripts or trends and highlight those to a human. What they’re really not good at doing, though, is taking the time to pull up YouTube and watching a video of a CEO who might look like he or she is hedging when asked a tough question in an interview. They don’t really demonstrate the ability to hone in on what I call BS factor. If we relied too much on these models, just to scrape data and words from preexisting text, and confuse that for true insight, I think we could stumble there. They haven’t really shown a true reasoning ability or reasoning faculty, but I think over time, they might get better at it. Companies like Meta are developing technologies that read human facial expressions in a minute fashion because they’re trying to create realistic avatars. Conceivably, that technology could be used down the road to map the human face. And maybe a large language model could merge up something like that and say, “Hey, Asit, there is a little BS factor going on in the CEO’s interview, you should pay attention.”
McCarty Carino: Do you feel like this kind of technology could replace what you do as an analyst?
Sharma: I still feel like I’ll have a job in the near future. The fun thing is that ChatGPT [and] large language models are evolving very rapidly. And our understanding of how to use them is also evolving. There are tons of new fun plug-ins that are coming into OpenAI, which puts out ChatGPT, as well as independent filters, front-end filters in the investment world. So I think this is going to be really great for those who want to experiment with the technology. I think it’s going to be a great helper to the way we evaluate stock recommendations. Myself, I don’t tend to outsource investment decisions to anyone, but this is what I do every day. I’ve got tons of friends who are very comfortable investing in [exchange-traded funds], where you just give your money to someone who’s managing a portfolio and you check on those results after some time, and I think that they’re naturally investors who will feel perfectly fine at some point in time as the models get better with asking ChatGPT for insights based on their preferences, building a portfolio and just checking in quarterly. So I don’t know about my job in the next few years. In the immediate future, maybe I can still come to work.
Related links: More insight from Meghan McCarty Carino
Recently the financial comparison site Finder did an experiment to test ChatGPT’s stock recommendations against several leading investment funds in the U.K.
The ChatGPT portfolio outperformed the other funds and the S&P 500 index over the same period.
Among the companies picked by ChatGPT for investment was Microsoft, which happens to be a major backer of ChatGPT’s maker, OpenAI, and has integrated the chatbot into its own products.
Which, I don’t know, seems like maybe a conflict of interest?
My Marketplace colleague Janet Nguyen also explored whether ChatGPT could be used to predict future monetary policy by analyzing Federal Reserve Chair Jerome Powell’s public comments. Overall, the chatbot was good at giving quick summaries of Powell’s speeches.
But when it comes to more nuanced interpretations, you still might want to tune in to “Marketplace” for that.
The future of this podcast starts with you.
Every day, the “Marketplace Tech” team demystifies the digital economy with stories that explore more than just Big Tech. We’re committed to covering topics that matter to you and the world around us, diving deep into how technology intersects with climate change, inequity, and disinformation.
As part of a nonprofit newsroom, we’re counting on listeners like you to keep this public service paywall-free and available to all.
Support “Marketplace Tech” in any amount today and become a partner in our mission.