LLMs as Simulations of Natural Language

We're joined by N. Katherine Hayles, Distinguished Research Professor in English at UCLA, to think through cognition in the broadest and most scaled sense. 

Hayles is among the foundational thinkers of posthumanism in its Anglophone register, and this conversation tracks her intellectual trajectory from the question of how we became posthuman to her most recent project: an integrated cognitive framework that extends from bacteria to AI. The opening provocation is one she has been developing since large language models appeared as a genuinely literary phenomenon, the claim that LLMs do not speak natural language but produce a computational simulation of it.

The umwelt of an LLM (its 'operative world-horizon,' in Uexküll's sense) overlaps with the human umwelt enough for communication to occur, but the divergences are large and consequential. This leads to the question of cognition itself. Against definitions that make consciousness the threshold of cognitive status, Hayles proposes the SIEPAL framework: Sensing, Interpreting, Responding, Anticipating, Learning, under which bacteria, algorithms, and ecosystems all qualify as cognitive. The non-conscious, on this account, isn't pre-cognitive but is in many ways more cognitively capable: faster, closer to environmental noise, less committed to the narratives of coherence that consciousness requires.

The final section breaks genuinely new ground with Hayles's turn to analog computation: the argument that digital computation is a historical blip, that biological life has always operated on analog principles, and that the future of computation (neuromorphic chips, organoid computers, hybrid analog-digital architectures) represents not a departure from but a return to what life has always done. She proposes the analog humanities as a corrective to digital humanities, and the computational humanities as the synthesis that might finally close the gap between biological and technological cognition. This one is very much worth enjoying in dialogue with our previous epsiode on the digital.

Some references:

N. Katherine Hayles

• How We Became Posthuman: Virtual Bodies in Cybernetics, Literature, and Informatics, University of Chicago Press, 1999
• Writing Machines, MIT Press, 2002
• Unthought: The Power of the Cognitive Nonconscious, University of Chicago Press, 2017
• Postprint: Books and Becoming Computational, Columbia University Press, 2021
• Bacteria to AI: Cognition Across Scales (referenced as new/recent book)

Leif Weatherby

• Language Machines: Cultural AI and the End of Remainder Humanism, University of Minnesota Press, 2025

Jakob von Uexküll — concept of the Umwelt; the species-specific world-horizon generated through particular sensory and neurological capacities

Walter Freeman

• How Brains Make Up Their Minds, Columbia University Press, 1999 — on EEG waves as the mediating mechanism between individual neurons and global hemispheric activation; the rabbit olfactory system experiments

Gregory Bateson — on systems that lose the ability to receive feedback collapsing; referenced without specific title (e.g. Steps to an Ecology of Mind, 1972)

Peter Haff — the technosphere

Stuart Kauffman & Giuseppe Longo, for arguing that biological organisms cannot be mapped into phase space and always follow the adjacent possible

Warren McCulloch & Walter Pitts — the McCulloch-Pitts neuron as a binary model with analog processes underlying the firing threshold

Bernd Ulmann — here referenced as an expert on analog computing who argues that continuity vs. discreteness is a secondary rather than primary distinction between analog and digital