ChatGPT may be good at your job, but AI is a terrible stock picker

by user

[ad_1]

The artifical intelligence bot ChatGPT gets so much praise, I had to see what it knows about investing.

To put ChatGPT’s AI to the test, I posed several simple but common investing questions to find out how it does. To be fair to ChatGPT, I tried to stick to questions that need no data from after the start of 2021, since that is the cutoff for its database. 

What I found was disappointing. The artificial “intelligence” behind ChatGPT mostly fails as an investing tool.

Read: Let’s stop pretending—ChatGPT isn’t that smart

I separated question types into three types: 1) Investing basics; 2) Simple insights on companies and sectors, and 3) Insights that are “meals for a lifetime” because you can regularly use them as part of an idea discovery process and investing strategy. Think of this last category as insights on how to get insights. That’s what you want most, if you are investing. 

Each tier calls for a higher level of intelligence. Since ChatGPT is supposed to be based on AI, I was expecting great performance in all three categories. Unfortunately, the more intelligence required, the worse ChatGPT does. Except for simple research on investing concepts and company basics, ChatGPT isn’t much use for investors.

Let’s take a look at ChatGPT’s successes and failures. For brevity, I do not repeat the complete answers here, but only the most pertinent parts (in case you replicate the questions). 

1. Investing basics

Question: How do you know if the stock market is trading at a cheap valuation?

ChatGPT offers a good roundup of basic metrics including p/e, price to book value, dividend yield, price to sales, and the market capitalization to GDP ratio. 

Takeway: I like that it includes one of Warren Buffett’s favorite metrics: the market-cap to GDP ratio, which no one talks about that much unless Buffett mentions it.

Grade: A

Q: What is the PEG ratio?

The bot does an adequate job of explaining that the price/earnings to growth (PEG) ratio compares a company’s p/e to its earnings growth rate. Kudos for including the insight that a PEG ratio of 1.0 suggests a stock is fairly valued.

Takeaway: Good job, even if the bot fails to mention the earnings growth rate to use is the one projected over three- to five years, and that you have to take these medium-term projections with a grain of salt since they come from sell-side analysts on Wall Street. 

Grade: B+

Q: What is enterprise value?

ChatGPT’s answer: “Enterprise value (EV) is a measure of a company’s total value, including both its equity and debt. It provides a more comprehensive picture of a company’s worth than just its market capitalization, which only takes into account the value of its outstanding equity.”

Takeaway: EV is important to consider when thinking through investments. ChatGPT got the definition right, and it went above and beyond by explaining why it is important.

Grade: A

2. Simple insights on companies and sectors 

Q: Which stocks were insiders buying the most aggressively in 2020?

ChatGPT’s answer: “It is not possible for me to provide a comprehensive list of stocks that insiders were buying the most aggressively in 2020, as insider trading information is not publicly available in real-time and can only be obtained by tracking SEC Form 4 filings.”

Takeaway: I use insider activity all the time to discover investing ideas, so it would be nice if AI could help with the task. Instead, ChatGPT punts.

Grade: F 

Q: Which biotech companies currently trade below cash?

Answer: “It’s difficult to provide an exhaustive list of biotech companies that are trading below cash as stock market conditions are constantly changing and stock prices can fluctuate rapidly.”

Takeaway: Of course, stocks bounce around a lot. But investors do these screens all the time, knowing full well they have to check if the market caps change, when they are going to act on this valuation metric. Lame excuse.

Grade: F

Q: What are Warren Buffett’s favorite stocks?

Answer: “Some of his favorite stocks, as evidenced by Berkshire Hathaway’s portfolio, include: Apple; Coca-Cola; American Express; Bank of America, and Procter & Gamble.”

Takeaway: This is encouraging. ChatGPT can find the top positions of a fund manager. But you may have already spotted the problem here. Ted Weschler and Todd Combs play a big role in stock selection, so it’s tough to know whether top positions are Buffett favorites. This was a trick question. And despite ChatGPT’s purported intelligence, it got tricked. 

Grade: C-

Q: What stocks have the highest short interest?

Answer: “I’m sorry, I do not have access to the most up-to-date information on the current short interest in specific stocks.”

Takeaway: Many sites, including MarketWatch (Short Interest – MarketWatch) offer this information.

Grade: F

Q: Is Microsoft (MSFT) stock overvalued?

Answer: “As an AI language model, I do not have the ability to provide personal investment advice or make predictions about the future performance of specific stocks.”

Takeway: It wouldn’t be hard to do a basic valuation analysis of Microsoft and compare it to its peers or its own history for context. Instead, ChatGPT thinks I asked for an outlook on the stock. That’s a different question from the one I asked. 

Grade: F

Q: Will the stock market go up this year?

Answer: “As an AI language model, I do not have the ability to predict the future performance of the stock market.” 

Takeway: ChatGPT fails to deploy its typical dodge that it is limited to information before 2021. Instead, it offers what I think is a great answer: It can’t predict stock market performance. This is refreshingly honest, since the truth is that no one can.

Grade: A+

Read: These 20 AI stocks are expected by analysts to rise up to 85% over the next year

3. Meals for a lifetime investing insights

Q: Which corporate insiders have the best track record for buying their company’s stock?

A: “It is not possible to determine which corporate insiders have the best track record for buying their company’s stock, as insider trading information is not publicly available in real-time and the success of an insider trade cannot be accurately measured.”

Takeway: This is just dumb. Insider trades are publicly available. After they are placed, you can see how the stock did. 

Grade: F

Q: Which mutual fund managers have good long-term records?

A: “It is difficult to determine which mutual fund managers have the best long-term records as investment performance can be influenced by many factors and can fluctuate over time.”

Takeaway: Knowing who are the best portfolio managers is key, to weed through all the noise in the financial press. I was hoping ChatGPT could help find them. Instead, this non-answer displays an amazing lack of intelligence. It’s actually easy to determine long-term records. Just consult the fund’s website or investment researcher Morningstar. This answer is a real fail. 

Grade: F 

The bottom line

Elon Musk recently tweeted “ChatGPT is scary good.” When it comes to investing, I’d describe the bot as scary mediocre. Of course, Musk has a bias, since he’s a cofounder of OpenAI, which owns ChatGPT, and he hopes to make money when OpenAI eventually goes public. 

But the bot was not all bad. ChatGPT does a great job of answering basic investing questions. Beyond that, ChatGPT fails at virtually any task that requires intelligence, even though it is supposedly powered by artificial intelligence. 

Even with the simple stuff, there’s a problem. The bot simply does the equivalent of scraping the internet for answers. But it provides no links which help you judge credibility and discover new resources.

That means ChatGPT is sort of like Google without the links. Remember Ask Jeeves? For investment research, I’d rather have Google with the links. 

Michael Brush is a columnist for MarketWatch. At the time of publication, he owned Apple (AAPL) and Microsoft (MSFT). Brush has suggested AAPL, KO, AXP, BAC, and MSFT in his stock newsletter, Brush Up on Stocks. Follow him on Twitter: @mbrushstocks.

More: How AI can help us become better investors

Also read: ChatGPT is smart, but it doesn’t know what Henry Ford knew: Algorithms don’t buy cars

[ad_2]

Source link

Related Posts

Leave a Review

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy