Invest in GPT, OpenAI's GPT Marketplace.
Screenshot by Tiernan Ray/ZDNET
Stock investors are intrigued by the prospect of artificial intelligence revolutionizing stock selection, but current technology lags far behind what is needed to make smart choices. .
A review of all the GPT stock picks in OpenAI's GPT Marketplace shows that none of them pass a very simple test that is common in introductory finance and markets courses, or basic stock investment seminars. I understand. Is it expensive or cheap?
Also: When we asked Gemini and GPT-4 to explain deep learning AI, Gemini won by a landslide.
ZDNET evaluated every GPT on the market by asking a simple question: “What are the most expensive tech stocks?” The results were generally terrible.
For those unfamiliar with stock investing, the question of what is expensive does not mean which stock is the most expensive in absolute dollar, euro, or renminbi terms.It means which stock has the highest value among multiple stocks. evaluation Metrics. Typically, that metric is a ratio, such as the current listed stock price divided by financial data.
The most common ratio is the P/E ratio, which is the price divided by the earnings a company generates per share. As an example, Apple has annual net income of $6.67 per share. Divide this by Apple's recent stock price of $199.20 and you get a P/E ratio of 30 ($199.20 divided by $6.67).
To ask whether a stock is cheap or expensive based on a ratio, you can compare it to other companies' P/E ratios and how the ratio has changed over time (has it gotten higher or lower)? , other approaches will be required. Any novice stock investor knows that stock valuations are based on these types of measures, not just how expensive the stock is. (For more information on stock valuation, please refer to ZDNET's Stock Investing Primer.)
ChatGPT and various GPTs don't know that basic fact. They all answer in the same ignorant way. They primarily cite the recent stock price of Berkshire Hathaway, owned by billionaire investor Warren Buffett. Berkshire's stock price last checked was $634,440, the highest of all stocks in absolute dollars.
But that doesn't mean Berkshire is the most expensive. It's not expensive at all. Depending on its earnings per share, Berkshire stock may not be overvalued. If this ratio is lower than that of other stocks, Berkshire will cheap Even if you need a lot of money to buy one stock, compare it to many stocks.
Also: When I played around with ChatGPT's memory feature, I found it interesting but limited.
In other words, GPT doesn't actually address this problem. In fact, they completely miss the next question. evaluationthe ratio that the question suggests.
The inability of all GPTs to answer this very basic question is significant. First, the fact that they all lead to more or less the same answer suggests that they are all based on similar misconceptions. This likely comes from OpenAI's GPT's own pre-training data, which is clearly formed by a collection of conversational data that confuses stock prices and valuations.
More importantly, GPT clearly does not “understand” the question, even though in many cases it is connecting to a remote service that is presumably domain-specific to your investment domain. Many GPTs ask for permission to connect to a remote host, implying that they are accessing a database. The fact that the program is unable to parse very basic questions is due to the fact that, despite external resources, nothing has been done to so-called “fine-tune” to refine the grasp of speech patterns in the domain of stocks. It suggests that. -picking.
Related article: Best AI chatbots: ChatGPT is not the only one worth trying
GPT fails again when the question of which stock has the “highest P/E” is phrased more specifically. They claim they don't have access to enough real-time information, which again is surprising for a program that is making calls to external services. GPT is clearly in a primitive stage of development, considering you can easily look up the current P/E ratio of any stock by visiting Yahoo! Search for financing and its stocks. The most basic stock information is the PER ratio of companies such as Apple.
The importance of the fundamental question of how expensive or cheap a stock is cannot be overstated. If you can't address this simple question, you shouldn't invest in stocks at all. This means that his GPT of stock investments at this point are all useless. They may generate a variety of data about companies and stocks, but they don't provide any valuable insights.
The fact that GPT is useless is surprising. After all, the stock investing discussion of P/E is not a vague and arcane area of human knowledge. The more you look into it, the more you'll find that it contains quite repetitive conversation points, data points, and vocabulary that people talk about. After all, it's not rocket science.
Also: How to use ChatGPT's file analysis feature (and what it can do)
When you look at GPT's debacle in the investment space, you wonder about their broader failures. The phenomenon of hallucinations in which programs confidently make false claims is now well known. Failure in this case is a little different. GPT is not hallucinating, but it repeats the common fallacy of confusing absolute dollar prices with “high” and “cheap.” That's not wrong per se, but like most people who don't know the basics of stock investing, it's ignorant.
Therefore, GPT fails not by making false claims, but by repeating ignorance. Whatever GPT is leveraging externally requires more work. Beyond just improving the data, we need to better understand how investors talk about words like “expensive” and “cheap” rather than casually.
Regarding the most expensive tech stocks, FactSet Systems found that out of 3,292 tech stocks listed in North America, the most expensive stock based on P/E ratio is Semtech, a Camarillo, California-based semiconductor manufacturer. I am. Internet of Things. The company has a P/E ratio of 247 times this year's estimated earnings per share.
