The AI wars have just begun and one of the largest financial giants in the world, Bloomberg, has reported the creation of their own artificial intelligence (AI) tool, called BloombergGPT.
What is BloombergGPT?
It is similar to the goal of OpenAI’s ChatGPT – revolutionize how people work. However, Bloomberg focuses on financial data and media business. The company has introduced a new large-scale AI model that aims to change the financial industry forever.
According to a Bloomberg release, BloombergGPT is a “large language model” (LLM) that has been trained on financial data to facilitate a wide range of natural language processing (NLP) operations in the financial sector.
That is a total game-changer for the financial industry. This implies that users may swiftly analyze financial data with the help of cutting-edge machine learning (ML) software, which can then be used for risk assessment, sentiment analysis, news categorization, etc.
“BloombergGPT outperforms similarly-sized open models on financial NLP tasks by significant margins — without sacrificing performance on general LLM benchmarks,” the announcement explained.
Bloomberg claims that by employing BloombergGPT, the entire potential of artificial intelligence will be brought to the financial industry. As a result, new opportunities will arise for using the massive amounts of data available on the Bloomberg terminal to better assist the firm’s customers.
Investors, bankers, and other finance executives use the Bloomberg terminal to obtain information and research about markets, companies, and industries in real time.
Bloomberg’s CTO, Shawn Edwards, sees an enormous value in having developed the first LLM targeted at the financial domain due to the following reasons: few-shot learning, text creation, conversational systems, etc.
“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,” he added.
By unleashing the full power of AI in the financial realm, BloombergGPT will open up new options for the company. It will be able to better assist its customers by marshalling the huge volumes of data available on the Bloomberg Terminal. It’s worth adding that Bloomberg stores decades’ worth of data.
The company mined this massive repository of financial information for documents in English to compile a 363 billion token dataset. A public dataset containing 345 billion tokens was also added, bringing the total number of tokens in the training set to almost 700 billion.
The team used a subset of this dataset to train a causal language model with 50 billion parameters for the decoder alone.
Bloomberg table chart, source: bloomberg.com