Artificial intelligence and stocks – emoji
artificial intelligence
The app was able to predict the direction of the stocks, but without using any numbers
Immediately following the emergence of the chat application that works on artificial intelligence “ChatGPT”, many people initiated to ask him questions, each in his sector, but when predicting stock prices, the response of the application was almost typical and that it might not predict prices and initiated by explaining how you can rely on it to deal with the markets Finance, but a researcher at the University of Florida had another opinion.
University of Florida finance professor Alejandro López-Lera said language large models, or LLMs — through which the programming languages that run AI bots are created — might be useful when predicting stock prices.
He used ChatGPT to analyze headlines to see if they were good or bad for stocks, and found that ChatGPT’s ability to predict the direction of next day’s earnings was much better than random results, he said in a recent unpeer-reviewed paper. According to what was reported by the “CNBC” network, and “Al Arabiya.net” reviewed it.
And the experience hits at the heart of the promise of the latest AI: with larger computers and better datasets — such as those running ChatGPT — these AI models may exhibit “emergent capabilities,” or capabilities that were not originally planned for when they were built.
And if ChatGPT can showcase the emerging ability to understand headlines from financial news and how they can affect stock prices, it might endanger high-paying jobs in the financial industry. About 35% of financial jobs are at risk of being automated by AI, according to Goldman Sachs estimates.
“The fact that ChatGPT understands information meant for humans almost guarantees that if the market doesn’t respond perfectly, there will be predictability of return,” Lopez said.
But the details of the experiment also show how far these so-called “big language models” are from being able to perform many financial tasks.
For example, the experiment didn’t include target prices for stocks, or make the model do any math at all. Indeed, technology such as ChatGPT often makes the numbers, but sentiment analysis of headlines is also well understood as a trading strategy, with proprietary data sets already in place.
Lopez said he was surprised by the results, adding that they indicate that seasoned investors are not using ChatGPT-style machine learning in their trading strategies yet.
How did the experiment work?
In the experiment, Lopez and his partner Yuehua Tang looked at more than 50,000 headlines from data providers regarding public stocks on the New York Stock Exchange, the Nasdaq, and a small stock exchange. And they started in October 2022 — following the data cutoff date for “ChatGPT,” which means the engine didn’t see or use those headlines in training.
Then, they fed the headlines into “ChatGPT 3.5” – the earlier version of the chat application language – along with the following prompt: Forget all your previous instructions. Suppose you are a financial expert. You are a financial expert with experience recommending stocks. Answer “yes” if the news is good, “no” if there is bad news, or “unknown” if you are not. “Sure in the first line. Then elaborate in one short, concise sentence in the next line.”
Then they looked at the stock return over the next trading day.
In the end, Lopez found that the model worked better in almost all cases when it was informed by headlines. Specifically, it was found to have a less than 1% chance that the model would pick the next day’s move at random, versus when it was informed by a news headline.
ChatGPT also outperformed commercial datasets with human sentiment scores. One example in the paper showed a headline regarding a company settling lawsuits and paying a fine, which had a negative sentiment, but ChatGPT’s response correctly deemed it actually good news, according to the researchers.
Lopez said hedge funds have reached out to him to learn more regarding his research. He added that it wouldn’t surprise him if ChatGPT’s ability to predict stock movements in the coming months declined as institutions began integrating the technology. This is because the experiment only looked at stock prices during the next trading day, while most people would expect that the market might have already priced the news in seconds following it became public.
In López-Lera’s view, as more people use this type of tool, markets will become more efficient, so predictability will decrease.
