Experts Urge Caution About Using ChatGPT To Pick Stocks 27
An anonymous reader quotes a report from Ars Technica: With AI chatbots growing in popular usage, it was only a matter of time before large numbers of people began applying them to the stock market. In fact, at least 1 in 10 retail investors now consult ChatGPT or other AI chatbots for stock-picking advice, according to a Reuters report published Thursday. Data from a survey by trading platform eToro of 11,000 retail investors worldwide suggests that 13 percent of individual investors already use AI tools like ChatGPT or Google's Gemini for stock selection, while about half say they would consider using these tools for portfolio decisions.
Unlike algorithmic trading, where computers automatically execute thousands of trades per second, investors are using ChatGPT as an advisory tool in place of human experts. They type questions, read the AI model's analysis, and then manually decide whether to place trades through their brokers. Reuters spoke with Jeremy Leung, who analyzed companies for investment bank UBS for almost two decades. Leung now relies on ChatGPT for his multi-asset portfolio. "I no longer have the luxury of a Bloomberg terminal, or those kinds of market-data services which are very, very expensive," Leung told Reuters. "Even the simple ChatGPT tool can do a lot and replicate a lot of the workflows that I used to do."
Reuters reports that financial products comparison website Finder asked ChatGPT in March 2023 to select stocks from high-quality businesses based on criteria like debt levels and sustained growth. Since then, the resulting 38-stock portfolio has reportedly grown in value nearly 55 percent. That performance beat the average of the UK's 10 most popular funds by almost 19 percentage points. But there's a huge caveat to that kind of AI success story: US stocks sit near record highs, Reuters notes, with the S&P 500 index up 13 percent this year after surging 23 percent last year. Those are conditions that can make almost any stock-picking strategy look smart.
Reuters frames the AI trading advice trend as a case of new technology tools "democratizing," or opening up, investment analysis once reserved for institutional investors with expensive data terminals. But experts warn that AI models can confabulate financial data and lack access to real-time market information, making them risky substitutes for professional advice. "AI models can be brilliant," Dan Moczulski, UK managing director at eToro, told Reuters. "The risk comes when people treat generic models like ChatGPT or Gemini as crystal balls." He noted that general AI models "can misquote figures and dates, lean too hard on a pre-established narrative, and overly rely on past price action to attempt to predict the future."
Unlike algorithmic trading, where computers automatically execute thousands of trades per second, investors are using ChatGPT as an advisory tool in place of human experts. They type questions, read the AI model's analysis, and then manually decide whether to place trades through their brokers. Reuters spoke with Jeremy Leung, who analyzed companies for investment bank UBS for almost two decades. Leung now relies on ChatGPT for his multi-asset portfolio. "I no longer have the luxury of a Bloomberg terminal, or those kinds of market-data services which are very, very expensive," Leung told Reuters. "Even the simple ChatGPT tool can do a lot and replicate a lot of the workflows that I used to do."
Reuters reports that financial products comparison website Finder asked ChatGPT in March 2023 to select stocks from high-quality businesses based on criteria like debt levels and sustained growth. Since then, the resulting 38-stock portfolio has reportedly grown in value nearly 55 percent. That performance beat the average of the UK's 10 most popular funds by almost 19 percentage points. But there's a huge caveat to that kind of AI success story: US stocks sit near record highs, Reuters notes, with the S&P 500 index up 13 percent this year after surging 23 percent last year. Those are conditions that can make almost any stock-picking strategy look smart.
Reuters frames the AI trading advice trend as a case of new technology tools "democratizing," or opening up, investment analysis once reserved for institutional investors with expensive data terminals. But experts warn that AI models can confabulate financial data and lack access to real-time market information, making them risky substitutes for professional advice. "AI models can be brilliant," Dan Moczulski, UK managing director at eToro, told Reuters. "The risk comes when people treat generic models like ChatGPT or Gemini as crystal balls." He noted that general AI models "can misquote figures and dates, lean too hard on a pre-established narrative, and overly rely on past price action to attempt to predict the future."
Re: (Score:2)
This post was indeed brought to you by ChatGPT.
Outsourcing stupidity (Score:2)
I wonder how many people have lost money betting after using ChatGPT to recommend what to bet on. Using it to pick stocks are in the same category of stupid.
Re: (Score:2)
While I generally don't care for AI LLMs these days, I don't suppose taking stock tips from it is any worse then following all these FOMO articles about the same topic to be any better.
Re: (Score:2)
There are uses for it. For example, say Trump decides he's going to impose tariffs on X countries, and you expect those countries to reroute their sales to the US (e.g. as knockdown kits, etc) through Y countries. For each of X, you can use AI tools to help find candidate stocks that would be most hurt by this event to short, and for each of Y, candidate stocks that would most benefit by this event to go long. You then manually investigate from the candidates to decide what you want your actual position
So it Matched the S&P 500? (Score:2)
The S&P 500 went from around 4,000 to over 6,000 since March 2023. So it replicated an index fund...
Re: (Score:2)
Don't forget dividend reinvestment, which is putting of total return.
People saw artificial intelligence in movies... (Score:3)
...so they think that today's AI is basically as intelligent as what they saw in the movies.
Spoiler alert: It's not. At all.
Sometimes I feel fortunate that my first few interactions with ChatGPT were marred by hallucinations. People who didn't get disappointed early are just going to be disappointed later. Possibly after trusting it with a decision that has real consequences.
Re: (Score:2)
As I have followed AI developments and research for nearly 40 years now, I had no expectations to begin with. I was somewhat impressed by the NLP interface, but my first question to ChatGPT was what its limits are and it nicely gave me an accurate list. Have occasionally used it for searching stuff since, with mixed results, especially when I asked for references.
3 too many words (Score:2)
Leave off the last 3.
Re:3 too many words (Score:4, Interesting)
Experts Urge Caution About Using ChatGPT To Pick Stocks
Leave off the last 3.
ChatGPT is decent in analyzing stocks in terms of underlying corporate financial or competitive fundamentals. The problem is that such analysis is no better at picking stocks than human analysis. The stock market is not bound by fundamentals. That's why game theory is a thing. This is not a knock on ChatGPT. It's a knock on stock buyers who follow the recommendations from other people or machines instead of understanding the underlying company.
I urge the opposite (Score:5, Funny)
As an investor who often makes money betting against retarded traders, I urge more retail investors to use ChatGPT. In fact, I urge them to use Claude or something to build custom HFT apps that rely on ChatGPT.
It's going to use training data (Score:2)
Re: (Score:2)
It is also not intended to give stock advice so it will be naturally bad at it and more likely to hallucinate advice.
True, but what is interesting is that it seems to be competing well with human stock advisors which perhaps says just as much about their methods too.
Re: (Score:2)
This is not interesting. Humans are notoriously bad at picking stocks. This is because the stock market is a random walk. You can't use data to predict future prices. That's why the best and most efficacious investment strategy is to buy a diverse portfolio of stocks, diversify into real estate holdings, bonds, commodities, and precious metals. (Whether you invest more in bonds or real estate depends on what the market is doing at the moment.) Preferably through a low fee index fund. That's more or less how
Uh (Score:2)
It's a large language model. It just does text.
That's extremely useful for many things. Picking stocks is not one of them.
If those "Experts" knew how to pick stocks best... (Score:4, Insightful)
The truth is, those "Experts" are as bad as dart-throwing monkeys or ChatGPT at picking stocks, but they earn money by making other people believe they knew something special to make them rich.
Re: (Score:2)
That's not true. Hedge funds are better than that because they can cheat at institutional scale.
Reverse AI psychology (Score:2)
Ask it about the best stocks to pick. Then ignore the ones it gives you and invest in others. Just because everyone else would pick the recommended ones, which would make it drop in value.
Re: (Score:2)
That's not how stocks work. Everyone buying them would lessen the supply while the demand was increasing, driving up the stock price. That's how the "pump" half of a pump & dump works.
AI is horrible at picking stocks. (Score:2)
There are 3 ways to beat the random guy off the street picking stocks:
1) Obtain information that is not public. I would be shocked if any publicly available AI had access to such information. This is why insider trading is illegal.
2) Manage your risk/return ratios well. This tends to require professional experience, and might involve things only the wealthy should do such as using long term vertical spreads.
3) Pick a low priced index fund - being sure to select an appropriate index (Nasdaq Composite is
Re: (Score:1)
Might not be so bad. (Score:2)
So if 1000 day traders ask chatgpt 1000 times... (Score:1)
for stock advice, will it give them the same answer?
Asking for a friend.
Re: (Score:2)
Indeed. We may finally have found a way to partially predict stock behavior.
Amazon (Score:2)
Go look at the number of book written about AI Stock Trading on Amazon and let me know if this is a scam.