Consider an analogy。 You are a manager， you have an employee， and you are trying to get a task done。 It helps to understand how the employee thinks， what the employee knows， and how they may behave。 What would happen if you made a certain request？ How would the employee interpret it？ How would they respond？ What do our past interactions with the employee tell us about their capabilities？ How should we phrase our request so that it is interpreted correctly？ Should we break our request into steps？ What aspects of their work would we need to verify？

Now imagine the employee is an AI。 It still helps to have answers for the questions above。 The goal of this book is to help you arrive at those answers。

We describe how a Large Language Model （LLM） operates in three different via input-output examples， via an overview of its training process， and via analogies to other familiar concepts。

You don't have to work with technology to understand this book。 It will be accessible to anyone with a high-school education—students， professionals， business-people， executives—anyone who wants to extract value out of LLMs or chatbots。

Author Mukund Sundararajan is a Distinguished Research Scientist at Google DeepMind。 He has spent the past decade analyzing， developing， and deploying AI in products—successfully and unsuccessfully。 He has also observed others do so。 He currently works on the Gemini series of Large Language Models and writes prompts for a living。