From the course: Agentic AI Solution Design Patterns
Join today to access over 25,300 courses taught by industry experts.
Internal monologue and behavioral design patterns
From the course: Agentic AI Solution Design Patterns
Internal monologue and behavioral design patterns
- In the previous video, we explained the foundational concept of internal monologue and the role of large language models and the controller module, as well as the use of system prompts. This first set of design patterns provides essential building blocks for how agents can leverage LLM driven reasoning for sophisticated planning, acting, and learning. In each of these pattern descriptions, we'll explain how the LLM and the controller module interact. By learning these design patterns, you will begin to understand how AI agents think and act. Before we begin, let's very briefly talk a bit more about system prompts. When the LLM is first activated to carry out a task, the agent can give it initial system prompts that tell it what to do. The first two upcoming design patterns, Chain of Thought and ReAct, are often applied via these initial system prompts that tell the LLM ahead of time to carry out a form of internal monologue suitable for the task at hand. But the agent can also issue…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.