welcome

Thank you for purchasing the MEAP version of “LLMs in Action”!

Coming from a background in data management and data science, I started looking into language models a few years ago. Back then, language models were considered an exotic topic in my community and few colleagues had worked with them. When giving talks on the topic, I would spend five to ten minutes introducing the basics of language models to my audience.

Fast-forward to 2023 and language models have taken the world by storm. ChatGPT, the fastest-growing consumer app in history, is nowadays known to and used by large shares of the world’s population, extending far beyond the computer science community. My introduction to language models has been reduced to a few slides that I nowadays often skip. Clearly, language models have become mainstream!

While many people have started using ChatGPT and similar models in their daily lives, I found that few people exploit their true potential. This is particularly true in the domain of data science. Perhaps for the first time ever, language models provide a unified interface that allows analyzing almost any kind of data, merely based on short instructions in plain English! But doing so efficiently and effectively requires using the right tools and interfaces, understanding the cost and quality tradeoffs between different models and configurations, and, ultimately, having a little bit of background knowledge of the internals of language models.