Module 1: Fundamentals of Generative AI#
This first module builds the foundation for everything that follows. By the end of it you will understand what foundation models and large language models are, how the transformer architecture made them possible, how to steer them with careful prompting, and how the same ideas extend from text into images and other modalities. Throughout, you will see how these capabilities are delivered as a managed service through Amazon Bedrock.
The module is organized as a continuous arc. We start broad, with the idea of a model that can generate new content, and progressively sharpen the picture:
No. |
Chapter |
What you will learn |
|---|---|---|
1 |
What generative AI is, where LLMs are used, and how Amazon Bedrock exposes foundation models through a single API. |
|
2 |
How foundation models differ from traditional ML, the transformer architecture and attention, and the practical limits of LLMs. |
|
3 |
The anatomy of a prompt, inference parameters, best practices, and in-context (zero-, one-, and few-shot) learning. |
|
4 |
Chain-of-thought, self-consistency, and tree-of-thought prompting for multi-step reasoning. |
|
5 |
Multimodal LLMs, how to prompt with images, and visual use cases. |
Each chapter ends by pointing you to the labs in Module 1 Labs: Hands-on with Amazon Bedrock, where the concepts become working Bedrock code.
How the pieces connect#
Generative AI sits on top of foundation models: large models pre-trained on vast data that can be adapted to many tasks. Large language models are foundation models trained on text. The transformer is the architecture that made today’s LLMs practical. Once you have a capable model, prompt engineering is how you direct it, and advanced prompting squeezes reliable reasoning out of it. Finally, multimodal models extend the same machinery beyond text to images and more. Keep this chain in mind as you read; every chapter adds one link.