BloombergGPT: data and media company introduces finance-focused AI model

The large language model will use financial data to support a diverse set of natural language processing tasks within the financial industry

Michael Bloomberg, the billionaire owner of Bloomberg LP and former New York mayor. AP

Financial data and media company Bloomberg introduced a new large-scale artificial intelligence (AI) model that aims to shake up the finance industry, much like OpenAI's ChatGPT programmes are seeking to transform written communication.

BloombergGPT is a large language model (LLM) that has been specifically trained on financial data to support a diverse set of natural language processing (NLP) tasks within the financial industry, Bloomberg said in a statement.

This means that the advanced machine learning software can quickly analyse financial data to help users make risk assessments, evaluate financial sentiment, classify news and get questions answered.

BloombergGPT will unlock new opportunities for using the vast quantities of data available on the Bloomberg terminal to better help the firm’s customers, while bringing the full potential of AI to the financial industry, the company said.

The Bloomberg terminal, which compiles real-time financial news, data and research reports about markets, companies and industries is used by investors, bankers and other finance executives to access information and research.

Bloomberg was founded by Michael Bloomberg after he worked for more than a decade and half on Wall Street for investment bank Salomon Brothers. The billionaire businessman served three terms as Mayor of the City of New York.

“For all the reasons generative LLMs are attractive – few-shot learning, text generation, conversational systems, etc – we see tremendous value in having developed the first LLM focused on the financial domain,” Shawn Edwards, Bloomberg’s chief technology officer, said.

“BloombergGPT will enable us to tackle many new types of applications, while it delivers much higher performance out-of-the-box than custom models for each application, at a faster time-to-market.”

The Organisation for Economic Co-operation and Development has urged governments, financial regulators and businesses to step up their efforts to address the challenges of developing and deploying trustworthy artificial intelligence in the financial sector.

AI, which is a set of applications including machine learning and robotics, holds “tremendous potential to improve productivity and innovation” in the financial sector, it said in its Business and Finance Outlook 2021 report.

However, its extensive use in the industry requires the right infrastructure that includes access to sufficient computational capacity and economical high-speed internet services, the Paris-based organisation said.

Bloomberg said it was seeking to use BloombergGPT to improve existing NLP workflows, while also finding new ways to put this model to work to serve its customers.

“The quality of machine learning and NLP models comes down to the data you put into them," Gideon Mann, head of Bloomberg’s ML Product and Research team, said.

"Thanks to the collection of financial documents Bloomberg has curated over four decades, we were able to carefully create a large and clean, domain-specific dataset to train an LLM that is best suited for financial use cases," he said.

The complexity and unique terminology of the financial sector require an AI model specific to financial data and BloombergGPT is a step in the development and application of this new technology for the financial industry, Bloomberg said.

To develop BloombergGPT, the company’s ML product and research group collaborated with its AI engineering team to create a comprehensive 363 billion token dataset consisting of English financial documents, drawing on the company’s existing data creation, collection and curation resources.

The engineers trained a 50-billion parameter decoder-only AI programme.

"Notably, the BloombergGPT model outperforms existing open models of a similar size on financial tasks by large margins, while still performing on par or better on general NLP benchmarks," according to Bloomberg.