That headlines can move markets is nothing new. Now two studies have tested how good ChatGTP is at classifying Fed texts and corporate news. Swiss banks are also probing their way forward with artificial intelligence.
A great number of pupils, students, and even those in the media have used artificial intelligence in the ChatGPT chatbot to generate texts. But how useful and reliable is the application of the text-based dialog system for the financial industry?
Two scientific studies have looked to answer that question. The results of bot analysis of Fed texts and company news were compared with the evaluation of market movements by analysts, according to a «Bloomberg» (behind paywall) story.
Nearly Human
The results show the prototype developed by the US company OpenAI has an edge over other AI-based systems. A study conducted by Fed employees, «Can ChatGPT Decipher Fedspeak?», concluded the system comes very close to human evaluation.
The aim was to evaluate whether the Fed's statements were more dovish or hawkish on monetary policy. Authors Anne Lundgaard Hansen and Sophia Kazinnik of the Richmond Fed concluded the chatbot beat Google's widely used «BERT» dictionary-based classification system.
Decoded Messages
In a second study, University of Florida researchers had ChatGPT act as if the bot were a financial analyst. The aim was to interpret headlines and news about companies according to whether they influenced the share price.
The results showed a statistical correlation with subsequent stock movements, showing how the bot was able to correctly analyze the impact of the news.
Skepticism in Switzerland
It should come as no surprise that in Switzerland, skepticism among independent asset managers about ChatGPT is high. According to a survey, more than half think text generated by ware based on AI is overrated. Accordingly, there is a certain reluctance to invest in corresponding AI companies.
Despite the reservations, customer advisors at banks are increasingly concerned about whether their profession will still exist in ten or twenty years because of artificial intelligence when banking will hardly require any human interaction with customers due to technical advances.
Trusting the Tech
Digital assistants need not be seen as competition to the traditional advisory business, however. Rather, they could become important productivity tools for the advisory business.
Critically, chatbots will only be used above all if they are useful and productive and the tech can be trusted.
Chatbot technology is also becoming increasingly important for internationally active Swiss banks. Those active in the Angloxon region and Asia, need to consider such tools, as financial institutions in these areas use banking chatbots much more frequently.
