The investment bank uses OpenAI’s ChatGPT to sniff out trading signals from Federal Reserve policy statements.
Investment banking giant JPMorgan has reportedly unveiled an artificial intelligence (AI) tool to analyze Federal Reserve statements and speeches to detect potential trading signals.
On April 27, Bloomberg reported the Wall Street investment bank is using a ChatGPT-based language model to digest comments from United States central bankers.
These Fed policy signals will be rated on a scale from easy to restrictive to derive what the bank has called a Hawk-Dove Score.
“Hawkish” is a monetary policy term that refers to raising interest rates to keep inflation under control. The opposite is “Dovish,” which favors an expansionary monetary policy and lower rates.
The AI tool will give analysts a way to detect policy shifts thatcould provide the bank with a heads-up on trading signals. “Preliminary applications are encouraging,” JPMorgan economist Joseph Lupton reportedly said.
The tool can be used to predict changes in central bank tightening. Hawkish policy statements for example could result in rising yields on one-year government bonds.
According to the JPMorgan model, which can analyze statements going back 25 years, Fed sentiment has fluctuated recently but remains predominantly hawkish.
The Federal Reserve is expected to raise its benchmark interest rate by another 25 basis points to 5.25% next week, according to Bloomberg.
A 10-point increase in the Hawk-Dove Score indicates a ten percent chance that there will be a rate hike at the next policy meeting and vice versa.
JPMorgan is keen on AI applications for its benefit but not so keen on letting its employees use them.
In February, the company restricted its staff from using ChatGPT, according to reports. No particular incident spurred the decision to block employees from accessing the AI chatbot and other firms have made similar moves.
In an annual letter to shareholders earlier this month, JPMorgan CEO Jamie Dimon revealed that the bank has over 300 AI use cases in production.