{"id":77243,"date":"2023-04-14T16:26:33","date_gmt":"2023-04-14T13:26:33","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=77243"},"modified":"2025-09-10T22:09:32","modified_gmt":"2025-09-10T19:09:32","slug":"bloomberg-unveils-bloomberggpt-a-large-language-model-for-the-financial-market","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/bloomberg-unveils-bloomberggpt-a-large-language-model-for-the-financial-market\/","title":{"rendered":"Bloomberg unveils BloombergGPT, a large language model for the financial market"},"content":{"rendered":"

The Bloomberg News Agency unveiled<\/a> the BloombergGPT large language model, capable of answering questions from the finance and business sectors.<\/p>\n

The 50-billion-parameter algorithm is built on the same technology as OpenAI’s GPT. According to the company, their neural network \u201cwith a significant lead\u201d outperforms open models of similar size in financial and NLP<\/span> tasks without sacrificing performance.<\/p>\n

\"Bloomberg
BloombergGPT test results in general and financial tasks. Data: Bloomberg.<\/figcaption><\/figure>\n

The specialists trained a LLM<\/span> on a corpus of 363 billion tokens of English-language financial data collected by the company\u2019s analysts over the last 40 years. They include Bloomberg\u2019s internal information, securities documents, press releases, news, and publications from various outlets.<\/p>\n

Experts also supplemented the training dataset with a general-purpose set of 345 billion tokens, containing information from GitHub, Wikipedia, YouTube and other sources.<\/p>\n

In training the Bloomberg model, Nvidia A100 accelerators were used in the AWS cloud. The training process took about 1.3 million GPU-hours.<\/p>\n

The model can determine whether news headlines are bearish or bullish for investors. It can replace company names with stock tickers, detect important names in documentation, and answer questions about business such as \u201cwho is the chief executive of the firm\u201d.<\/p>\n

The technology also has some generative AI capabilities. For example, it can write article headlines based on short annotations.<\/p>\n

The agency said it does not plan to release the algorithm due to the risk of leaking confidential data.<\/p>\n

The company also does not intend to compete with OpenAI’s ChatGPT. For now, Bloomberg is using AI to expand the functionality of its Terminal service for studying and analysing the financial market.<\/p>\n

Back in January 2023, it emerged that the information portal CNET had used artificial intelligence<\/a> to write a series of articles.<\/p>\n

Subsequently, readers and editors discovered errors in the texts. As a result, the publication paused the release of new materials.<\/p>\n

Nevertheless, two weeks later, CNET said that they would continue to write articles<\/a> with the help of AI.<\/p>\n","protected":false},"excerpt":{"rendered":"

Bloomberg News Agency unveiled the BloombergGPT large language model capable of answering questions from the finance and business sectors.<\/p>\n","protected":false},"author":1,"featured_media":77244,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"1","news_style_id":"1","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,1569,1201],"class_list":["post-77243","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-bloomberg","tag-chatbots"],"aioseo_notices":[],"amp_enabled":true,"views":"17","promo_type":"1","layout_type":"1","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/77243","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/comments?post=77243"}],"version-history":[{"count":1,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/77243\/revisions"}],"predecessor-version":[{"id":77245,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/77243\/revisions\/77245"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media\/77244"}],"wp:attachment":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media?parent=77243"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/categories?post=77243"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/tags?post=77243"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}