TL;DR:
- Bloomberg is incorporating AI similar to OpenAI’s GPT into its terminal software.
- Bloomberg GPT was specifically trained on financial data, unlike OpenAI’s GPT, which was trained on general text.
- Bloomberg GPT can answer business queries accurately and write headlines based on short blurbs.
- Large language models are becoming a key focus in the tech industry, with big players such as Microsoft and Google investing in them.
- Bloomberg’s GPT was developed using off-the-shelf AI methods and niche data, making it a cutting-edge AI model that does not rely solely on large investments or specialized resources.
- The model is highly versatile and can be used for a range of finance and business tasks, such as analyzing news headlines and identifying key individuals in a document.
- Bloomberg is exploring ways to utilize AI to help financial professionals manage the deluge of news stories, including summarization, monitoring, and question-answering capabilities.
- Bloomberg’s GPT is being integrated into the features and services accessible through its Terminal product.
Main AI News:
Bloomberg LP is taking on the AI industry with its own model, Bloomberg GPT, which utilizes the same underlying technology as OpenAI’s GPT. According to a company official, the model is set to be integrated into the features delivered through Bloomberg’s terminal software.
The benefits of Bloomberg GPT are already evident. It provides more accurate answers to questions like “CEO of Citigroup Inc?” and has the ability to assess whether headlines are bearish or bullish for investors. In addition, the model can even write headlines based on a short blurb, showcasing its versatility.
The development of large language models that are trained on terabytes of text data has become a key focus in the tech industry. Companies like Microsoft and Google are leading the race to integrate this technology into their products. Startups in the artificial intelligence space are also regularly raising funds at valuations over $1 billion.
Bloomberg’s move towards AI reflects how software developers in a range of industries are embracing state-of-the-art technology to automate tasks that were once performed exclusively by humans. Gideon Mann, head of ML Product and Research at Bloomberg, has stated that the capabilities of GPT-3 and its performance through language modeling were unexpected but have transformed the way natural language processing (NLP) is conducted at Bloomberg.
The move also signifies how the AI market may not be dominated solely by industry giants with extensive data reserves. With the development of innovative models like Bloomberg GPT, smaller players can establish their presence in this rapidly evolving industry.
The emergence of large language models has sparked concerns about the financial barriers to entry into the AI industry. With the need for access to supercomputers and millions of dollars, it is feared that OpenAI and Big Tech companies will gain an insurmountable advantage, selling access to their models to others and establishing themselves as the dominant players.
However, Bloomberg’s GPT has been developed without the use of OpenAI’s resources. The company has leveraged off-the-shelf AI methods and applied them to its proprietary, yet niche, data store. This approach has yielded promising results in a range of tasks, including the ability to determine the financial impact of a news headline on a company, changing company names to stock tickers, identifying key individuals in a document, and answering basic business queries such as the name of a company’s CEO.
Bloomberg’s GPT also has some generative AI capabilities, such as suggesting new headlines based on short paragraphs. While the company’s approach may differ from that of other players in the industry, the results speak for themselves. Bloomberg has shown that it is possible to develop a cutting-edge AI model without relying solely on large investments or access to specialized resources.
Bloomberg’s GPT is a highly specialized AI model that has been trained on a massive collection of financial documents, making it an ideal tool for the finance and business industries. While OpenAI’s GPT was designed to be a foundational model, Bloomberg’s approach is more targeted.
The model was created using a combination of proprietary and public data sets. About half of the data used to train the model comes from nonfinancial sources, such as GitHub, YouTube subtitles, and Wikipedia. The remainder of the data comes from FinPile, Bloomberg’s proprietary data set, which contains over 100 billion words of financial data accumulated over the past 20 years. This includes securities filings, press releases, Bloomberg News stories, stories from other publications, and a web crawl that focuses on financial web pages.
Bloomberg’s GPT is highly versatile and can be used for a range of tasks in the finance and business sectors. Its ability to analyze news headlines and assess their impact on a company’s financial outlook is just one example of how the model could be used. The model can also assist with tasks like identifying key individuals in a document and answering basic business queries.
Bloomberg’s decision to incorporate specific training materials into its GPT has resulted in increased accuracy and performance in financial tasks. As a result, the company is planning to integrate the model into features and services accessible through its Terminal product.
One of the model’s early applications will be to convert human language into the specific database language that Bloomberg’s software uses. For instance, it will translate “Tesla price” into “(get(px_last) for([‘TSLA US Equity’]).” Additionally, the model can perform behind-the-scenes work such as cleaning data and running other errands on the application’s back end.
Beyond these technical applications, Bloomberg is exploring ways to utilize AI to help financial professionals save time and stay informed about the latest news. The company is currently developing features that will help clients manage the deluge of news stories, including summarization, monitoring, and question-answering capabilities.
According to Gideon Mann, head of ML Product and Research at Bloomberg, there are numerous potential applications for the model. Bloomberg aims to leverage the power of AI to provide innovative solutions for financial professionals, ensuring that they can make informed decisions and remain competitive in the market.
Conlcusion:
Bloomberg is incorporating AI similar to OpenAI’s GPT into its terminal software, with the model being highly specialized in finance and business and trained on a massive collection of financial documents. The model can answer business queries accurately, write headlines, identify key individuals in a document, and assess the financial impact of news headlines on a company.
Bloomberg is exploring ways to use AI to help financial professionals manage the deluge of news stories and automate tasks that were once performed exclusively by humans. The move also signifies how the AI market may not be dominated solely by industry giants, as innovative models like Bloomberg GPT can establish their presence in this rapidly evolving industry.
