Abstract

The advent of Generative AI has reignited debates about the nature of creativity. Central to this discourse is the parallel between how humans draw inspiration from past art and how AI models utilize pre-training data. This essay argues that this parallel is not merely an analogy but a functional reality: the AI pre-training phase is the direct technological equivalent of human artistic inspiration, and creativity itself is best understood as a process of combinatorial synthesis. We will demonstrate that both human and machine rely on a vast corpus of prior work to synthesize novel creations. Consequently, we argue that restricting AI’s access to data during this crucial inspirational phase, a process we term "creative starvation", would not foster originality but instead lead to the technical and economic "death" of the generative AI industry, crippling its potential as a transformative tool for art, culture, and cognition.