Perception versus Training in AI Art

Distinguish two kinds of machine art: outputs produced by statistical recombination of human works (training‑based) and outputs produced by embodied systems that perceive and respond to the environment (perception‑based). This distinction matters for how audiences, institutions, and laws assign authorship, value, and moral standing to creations labelled 'AI art.' — If perception‑based systems become common, debates about copyright, authenticity, and whether machines can 'create' will shift from technical details to broad cultural and legal reassessments.