Imagine an AI that could pen a captivating novel, weaving intricate plots and developing believable characters. That's the ambitious goal of researchers behind a new approach called "Ex3," short for Extracting, Excelsior, and Expanding—and it's showing some promising results. Generating long-form narratives has always been a challenge for AI. Typical language models struggle to maintain coherence and consistency over thousands of words. Ex3 tackles this by mimicking how human authors learn: by reading and summarizing existing works. First, Ex3 "Extracts" structural information from a vast collection of novels, breaking them down into chapters and summarizing key plot points. This extracted knowledge forms a hierarchical structure, much like an author’s outline. Next, in the "Excelsior" stage, this structured data is used to create a specialized training dataset. A large language model (LLM) is then fine-tuned on this dataset, essentially teaching it the art of novel writing, including plot development, character arcs, and genre-specific language styles. Finally, the "Expanding" stage unleashes the AI's creativity. Given a short premise or seed idea, the LLM uses its learned knowledge to expand it into a full-fledged novel, chapter by chapter. It maintains coherence by tracking entities like characters and locations, ensuring consistent development throughout the story. Testing shows Ex3 outperforms other AI novel-writing methods in terms of plot coherence, engaging narratives, and human-like writing. While it's still under development, Ex3 represents an exciting step towards AI-authored fiction. However, it has limitations. Current testing is primarily with Chinese novels, further research is needed in other languages. Moreover, the initial premise or outline heavily influences the output's quality. Introducing feedback and revision mechanisms could further refine the stories and imbue them with more originality. But overall, Ex3 offers a tantalizing glimpse into a future where AI collaborators might help writers create more engaging narratives, perhaps even penning bestsellers themselves. This research could redefine how we think about creativity and storytelling in the digital age.