If you want to get a better understanding of generative AI, it pays to see what the New York Times and Bloomberg are up to

One of the problems with generative AI like ChatGPT is it makes you think it is magical. You type in some prompt, it comes back with a complex answer, the next thing you know, you are thinking this thing is smarter than a human. It doesn’t help that there is so much hype surrounding the technology. All of this can make you think it’s supernatural.

Well, it isn’t. It’s simply good IT. It consists of data, hardware and software, just like any other IT. To get a better appreciation of the ordinary nature of that, it helps to look at two recent examples: the AI the New York Times recently built and the AI Bloomberg just built.

It’s best to start with what the Times built. They used software called nanoGPT (karpathy/nanoGPT: The simplest, fastest repository for training/finetuning medium-sized GPTs) and took the works of Jane Austen, Shakespeare and more to build a chatGPT-like program on their laptops. Then they walked through the steps of getting it working, here: Let Us Show You How GPT Works — Using Jane Austen – The New York Times. It works pretty well after much much training. Obviously it is not as massive or sophisticated as ChatGPT, but after reading the article, you will have a better sense of how this technology works, and why it’s impressive but not magical.

After that, I recommend reading more about BloombergGPT. Their press release states:

Bloomberg today released a research paper detailing the development of BloombergGPT, a 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.

You can find a link to that research paper, here:  BloombergGPT: A Large Language Model for Finance. What I liked about that paper is it walks through the approach they took, the data they used, and the technology deployed to make their model. Even better, they talk about how it is currently succeeding and what some of the limits of it are.

I’m happy that both these companies have been good about sharing what they are doing with this technology. I might even try and use an old laptop to build my own AI. I mean who wouldn’t benefit from tapping into the genius of Shakespeare or Jane Austen.

For more on what Bloomberg is doing, see this: Bloomberg plans to integrate GPT-style A.I. into its terminal



Comments are closed.