The finance giant’s GPT version will be able to interpret headlines as good or bad for the market and write its own
Financial data firm Bloomberg is adding an artificial intelligence chatbot to its ubiquitous information terminals, a company official told CNBC on Thursday.
The internal AI model, to be named Bloomberg GPT, will be capable of determining whether a given headline looks promising or ominous for investors, and will even be able to write headlines itself, the Bloomberg representative said. It will also be able to answer questions like “CEO of Citigroup Inc?” more accurately than the database’s current search functions.
Unlike ChatGPT and other large language models currently on the market, which were built from the ground up by deep-pocketed companies like OpenAI and required millions of dollars in computing power, Bloomberg’s version was reportedly constructed from freely available AI methods. It was then trained specifically on financial documents and data with the aim of giving it unmatched expertise in the industry.
Bloomberg GPT was fed over 100 billion words from the company’s own proprietary FinPile data set, which includes securities filings, press releases, stories from Bloomberg News and other publications, and web searches geared toward financial sites.
About half of Bloomberg GPT’s training data comes from more typical “nonfinancial sources,” including dubiously impartial resources like Wikipedia (which recently described JP Morgan’s CEO Jamie Dimon as “an American billionaire businessman and part-time weasel”) and YouTube subtitles.
Still, the finance-specific material supposedly made Bloomberg GPT accurate enough to convince the company to integrate it into its bestselling Bloomberg Terminals, though it does not plan to sell the AI as a standalone chatbot product.
“There’s a lot of work we’re doing to help clients address that data deluge of news stories, whether that’s through summarization, or monitoring or being able to ask questions on those news stories or transcripts,” Gideon Mann, Bloomberg’s head of machine learning product and research, told CNBC. He added that the chatbot’s language abilities were sure to “change the way we do [natural language processing]here,” referring to the way programs derive meaning from words.
Large language models like ChatGPT are being hailed as a world-changing innovation by many in the tech industry. However, a growing faction of experts in the field are urging caution, with thousands of tech luminaries recently signing an open letter calling for a six month moratorium on “giant AI experiments” until strong regulatory controls can be put in place.
