Summary
In this chapter, we introduced how an LLM can be the brain of a sophisticated and complex system. We can use the conversational and reasoning abilities of the LLM to solve a task. As we said, this brain can be extended by providing perceptual systems (senses) and tools (hands). In fact, we can allow the model to search the internet by connecting with APIs, but also to ingest information from other modalities (audio, images, or video). Similarly, the model uses this received information to solve user tasks. If we can imagine agents performing and automating routine tasks for users today, it is not difficult to imagine a world in which agents interact with humans and other agents in increasingly sophisticated and complex ways.
In the next chapter, we will see how a model can have a memory, as well as how to store information and be able to find it again to be more efficient.