What kind of meaning can machines make — and why does it matter that it’s not the same as ours?
Paul Kockelman's Last Words (2024) offers a rigorous but accessible account of how large language models actually work — and why the meaning they produce is fundamentally different from human meaning-making. Drawing on the semiotics of C.S. Peirce, he shows that LLMs are trained to predict word-word relations, not word-world relations, which explains both their uncanny fluency and their systematic blind spots. The result is a compact, essential guide to cutting through the hype: not a dismissal of AI, but a precise account of what it can and cannot do — who profits from the confusion.
Stay tuned for Paul Kockelman’s expanded thoughts on Last Words.
Commentary from our contributors
The conversation elsewhere
Pundits, professors, and politicians are discussing The Know-Nothing Campaign and its concerns across media and the university ecosystem.