Engineering Creativity: The Quest for AI-Generated Literature pt. 1
A personal hypothesis is that the ideal AI text generator would need to account for literature having both a thematic axis (anthropology and worldview transformed into emergent symbols) and a stylistic one (command of linguistic code). The engineering has to be designed 'around' language models — because no amount of computing power alone will enable the engine to write with flair. The chassis, wheels, axles etc. all need to be engineered. The hope remains to feed the machine information and expect it to spontaneously generate a new G. R. R. Martin or the eighth Harry Potter. But I contend: given comprehensive data profiling an author's themes and style parameters — frequency of grammatical structures, morphological choices etc. — it may be possible to coach a text-handling sub-intelligence through countless corrections, paraphrases and micro-adjustments, steering it toward the desired product. An LLM may never itself be the tool that writes interesting books or movie scripts, but rather the driving force behind a machine capable of doing so.
Subscribe to my newsletter
Read articles from Joruje directly inside your inbox. Subscribe to the newsletter, and don't miss out.
Written by