It is not nerds, but scribes who can explain the field of generative artificial intelligence such as ChatGPT: such “large language models” have their roots in linguistics, literary criticism, and even the interpretation of sacred texts. Martin Warnke pursues this thesis through an archaeology of the origins of LLM – from Zellig S. Harris' statistical language theory to Walter Benjamin's philosophy of language to Jewish Kabbalah. In doing so, he shows that mystical text interpretation techniques have similar characteristics and thus also problems as today's computer systems: It is language structures, not artificial brains, that fascinate and frighten us. The “hallucinations” of AI are not teething problems, but the inevitable consequence of linguistic processes that derive semantics from syntax alone, without reference to the world. Criticism must therefore start with the fundamental question: What are these systems that exist only in language and have no world?
Essay
Martin Warnke, born in 1955, holds a doctorate in theoretical physics and was Professor of Computer Science and Digital Media at Leuphana University Lüneburg. There, he was director of the DFG Collaborative Research Group 'Media Cultures of Computer Simulation'.
By the same author(s)
"[...] the author, a physicist and historian of digital media, explains concisely and compellingly why large language models (LLMs) can produce a lot, but not intelligent behavior." – Günter Hack, FAZ
