LLMs as Cognitive Systems

N. Katherine Hayles is a professor of English at UCLA and Emeritus Professor of Literature at Duke University. She is also the author of a number of books on consciousness and AI. Her latest book is titled Bacteria to AI: Human Futures with Our Nonhuman Symbionts.

Greg and Katherine discuss technics - recursive feedback loops in which humans and tools co-evolve. Katherine argues that cognitive technologies and AI intensify this process, so we design them while they also design us. She distinguishes cognition from consciousness, emphasizing fast nonconscious neuronal processing and defining cognition as interpreting information in context with meaning, operationalized by SIRAL (sensing, interpreting, responding flexibly, anticipating, learning). 

Katherine claims plants and bacteria meet these criteria, while physical processes are agents without choices; cognitive systems are actors that select and adapt. She applies this to computation, treating deterministic mechanisms as noncognitive but viewing modern systems and LLMs as cognitive, discussing aboutness via biosemiotics and LLMs’ “conceptual environment.”

*unSILOed Podcast is produced by University FM.*

Episode Quotes:

Are humans and AI evolving toward each other?

07:29: So we can chart the evolution of humans and cognitive computational media in just this fashion. So humans start by being immersed in their environment. They could not survive otherwise. And then humans evolve up to abstraction. Computers start with abstraction, and now, with sensors and actuators and networking, they evolve toward immersion. So humans start with purpose. Their purpose is to survive. That's true of all biological organisms. And then they evolve up to design. Computers start with design. But now, with AI, they seem to be evolving toward purpose, which is the same as biological purpose, to survive. 

Consciousness is based on selfhood and self-narration

10:27: Consciousness is based on selfhood and self-narration. The stories we all tell ourselves every moment of every day about who we are and what we're doing, and that consciousness frequently lies. We know that eyewitness reports, for example, are often very untrustworthy because people just perceive what consciousness wants them to perceive. And often that is not accurate. One of the primary purposes of consciousness is to make the world make sense. When highly unusual phenomena happen, consciousness just edits it out.

AI can now see humans from the outside

37:23: So we're using our projective capabilities to imaginatively construct an umwelt and then seeing what that would mean for our existence, our sense of meaning or whatever. But we're always doing that from the outside. We're never inside anything but the human umwelt. Now we have a technology in large language models that is capable of seeing the human umwelt from the outside and telling us about it. That has never happened before.

Show Links:

Recommended Resources:

• Bernard Stiegler
• Inclusive fitness
• Chiasmus
• Consciousness
• Daniel Dennett
• John Searle
• Stochastic parrot
• Biosemiotics
• Umwelt
• Symbiosis
• Context window
• LLM
• Terrence Deacon

Guest Profile:

• Faculty Profile at UCLA
• Faculty Profile at Duke
• Wikipedia Profile

Guest Work:

• Amazon Author Page
• Bacteria to AI: Human Futures with Our Nonhuman Symbionts
• Postprint: Books and Becoming Computational
• The Cosmic Web: Scientific Field Models and Literary Strategies in the Twentieth Century
• Chaos Bound: Orderly Disorder in Contemporary Literature and Science
• Unthought: The Power of the Cognitive Nonconscious
• Chaos and Order: Complex Dynamics in Literature and Science
• How We Think: Digital Media and Contemporary Technogenesis
• My Mother Was a Computer: Digital Subjects and Literary Texts
• Electronic Literature: New Horizons for the Literary
• Writing Machines

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.