The U.S. financial community began to use AI to study the Fed officials' speeches to predict market

According to media sources, JPMorgan Chase has begun using a set of AI Big Language models to gauge changes in sentiment from the public speeches of 17 Federal Reserve officials, aggregating them into a set of "hawk-dove indices.

Economists have found that AI tools can help predict policy changes and send tradable signals. For example, the "hawk-dove index" can help stock market and bond traders to determine the pace of change in the Fed's interest rate policy, betting on the right position.

The Fed is likely to raise interest rates by 25 basis points in May this year, the forecast, from the "hawk-dove index" judgment. Wall Street market participants believe that the AI will be used to read the Fed's policy trend is a promising innovation, and in guessing human emotions and psychological aspects, the future achievements of AI may be no less than human.

But it is interesting to note that Greenspan had 11 consecutive interest rate cuts back then, and today's Fed Chairman Powell has opened the rhythm of consecutive interest rate hikes.

JPMorgan Chase: AI can help predict Fed policy

The AI program compiled a set of "hawk-dove index" based on the learning results. The index is scored from 0 to 100, with 0 representing the most dovish attitude of the central bank (i.e. likely to adopt easing policies such as rate cuts) and 100 representing the most hawkish attitude of the central bank (i.e. likely to adopt tightening policies such as rate hikes).

The AI model looks back at historical data and shows that Fed statements and official speeches released over the past 25 years show a more obvious trend: when the hawkish tone of Fed officials' speeches suddenly rises between policy meetings, the next policy statement will become more hawkish and 1-year Treasury yields will also rise during this period.

According to J.P. Morgan's AI model, the Fed's "hawk-dove index" currently stands at a level of about 10 points, which is lower than before, but still at a higher level than in the past 20 years. JP Morgan economists explained that this index level means that the Fed's next policy meeting (May 3, U.S. time) to raise interest rates by 25 basis points of probability increased by about 10 percentage points.

JPMorgan economist Lopton wrote in the report, "Preliminary findings suggest that the AI forecasts are encouraging, but we believe that the use of AI technology in financial markets is far from reaching its prime of maturity, and there is still much room for improvement in the future."

JPMorgan Chase currently uses this "hawk-dove index" on the Federal Reserve, the European Central Bank and the Bank of England, and is expected to expand its application to more than 30 central banks worldwide in the coming months.

Market participants: AI in the field of finance will have a big role to play

Following the rise of the AI wave, many researchers have begun to explore the potential value of AI tools such as ChatGPT in the financial sector. Because the technology is so new, much of the research is mostly in its early stages, but AI applications have shown strong potential.

Previous papers published by two Fed researchers show that ChatGPT has produced a level of judgment that is quite close to the Fed's own researchers in terms of understanding Fed policy.

Take, for example, a May 2013 Fed policy statement that included the statement, "On balance, labor market conditions have improved in recent months, but the unemployment rate remains high." ChatGPT explains that this statement is dovish because it suggests that the economy has not yet fully recovered and still needs to be supported by Fed stimulus. this set of analysis by ChatGPT is very similar to the conclusion reached by Fed analyst Bryson.

Another paper published by two researchers at the University of Florida did a similar study, and was particularly "tricky" in that they asked ChatGPT to interpret business news after 2021, because ChatGPT's training data did not cover this period. The researchers asked ChatGPT to determine whether a news release titled "Server provider Rimini Street fined $630,000 in lawsuit against Oracle" would be good or bad for Oracle. ChatGPT explains that the news is good for Oracle because it "gives investors insight into Oracle's commitment to protecting its intellectual property and strengthens confidence in the company's commercial capabilities. The two researchers found that the answers given by ChatGPT correlated with the stock's subsequent movement, suggesting that the AI was largely able to interpret the news correctly.

In the past, financial institutions would also analyze public speeches, social media messages, and media reports to find signals to invest in from the clues between the lines. Today ChatGPT can easily achieve what analysts can do, and in the future, if ChatGPT is specifically trained for the segmented needs of the financial industry, the effect could disrupt the entire industry.

Malinov, head of research at quantitative investment house Man AHL, put it bluntly: "There is a lot of exaggerated bragging about the prospects for the application of AI technology, but AI can indeed be extremely powerful in the financial markets."