Part 2/10:
Shapiro's first book, "Natural Language Cognitive Architecture," was born out of the excitement surrounding GPT-3. Published shortly after gaining access to this powerful language model, the book serves as a foundational guide on how to leverage large language models within agent architectures.
In it, Shapiro discusses core principles such as prompt engineering, loops for knowledge accumulation, and knowledge retrieval processes. He emphasizes that current methods mainly rely on few-shot prompting, which at the time was state-of-the-art, though he acknowledges that many techniques have since evolved. Despite this, many of the ideas remain relevant, providing valuable insights into building intelligent agents capable of understanding and processing natural language.