First, Bloomberg has announced BloombergGPT, new large-scale generative artificial intelligence (AI) model. This large language model (LLM) has been specifically trained on a wide range of financial data to support a diverse set of natural language processing (NLP) tasks within the financial industry.
The company says it was trained on a corpus of over 700 billion tokens (or word fragments). For context, GPT-3, released in 2020, was trained on about 500 billion. Of the 700 million-plus tokens, 363 billion are taken from Bloomberg’s financial data, the sort of information that powers its terminals — “the largest domain-specific dataset yet” constructed, it says. Another 345 billion tokens come from “general purpose datasets” obtained from elsewhere. The company-specific data, named FinPile, consists of “a range of English financial documents including news, filings, press releases, web-scraped financial documents, and social media drawn from the Bloomberg archives.”
Second, you may have seen the video of the woman in a black turtleneck who says you can’t trust your eyes. It’s an AI-generated deep fake; the video also tells on itself. A menu in the top left helpfully informs you that the video “contains AI-generated content.” There’s a timestamp and a credit. It was released on Tuesday by technology company Truepic and production studio Revel.ai to promote the idea of a relatively new transparency standard for digitally created content.
Why do we care?
I wanted to end the week with some big thoughts here. Chris Messina and Brian McCullough at the Tech Meme Ride home have offered a theory of varietals of AI. Like wine and the choice of grape, AI and the choice of model will produce different outcomes. I like this concept… because it extends nicely to the sommelier. Just like a sommelier can match a customer to the wine that fits their palate, a solution provider can be the sommelier for marching a customer to the AI model that fits their needs.
BloombergGPT is an example of how these models become specific for the industry, which is good for a diversity of ecosystems and options for service providers.
And I wanted to highlight the technology that may be part of the authentication of digitally created content because trust will be a crucial element to their use.