Why Computer AI Will Never Do What We Imagine It Can

ABSTRACT:In response to questions about the author’s previous proof (in Narrative 29.1) that computers contain a hardware limit that renders them permanently incapable of reading or writing narrative, this article draws upon the author’s work with Deep Neural Networks, Judea Pearl’s do-calculus, GPT-3, and other current-generation AI to logically demonstrate that no computer AI (quantum or otherwise) has ever learned, or will ever learn, to produce or process novels or any other kind of narrative (including scripts, short fiction, political speeches, business plans, scientific hypotheses, technology proposals, military strategies, and plots to take over the world).