When we use LLM AI models for procedural content generation, the underlying black box primarily harnesses statistical models to populate the gaps trying to be filled in. Its intrinsic function aims for mainstream appeal, meticulously curating outputs that deviate minimally, hence minimizing surprises.

This highlights an underlying tension (Zettelkasten edges spark learning and ideas), where utilizing a mainstream tool to create unique and surprising results presents a paradox. Engaging with these procedural tools requires navigating the push and pull between seeking originality and acquiescing to the predictability forced by the mass populace's training data (Safety Dance swimming story).

Source: Degrees of comfort, levels of control

#toolong