The Human Soul is LLM-like

As LessWrong contributor Julian Bradshaw explores in a recent essay, there's an intriguing parallel between a human soul and the weights inside a large language model: both are immaterial and conceptual, both define identity, and both are instantiated through physical hardware. That comparison raises genuinely philosophical questions: if a soul is just a pattern, what prevents multiple instantiations across time or space? And would surviving as only your mind, without your body and experiences, even constitute survival? Bradshaw suggests that understanding how large language models maintain identity through fine-tuning and versions might illuminate the ancient puzzle of how humans remain themselves while changing constantly. The essay circles back to a central ambiguity that has haunted soul-concepts for centuries: how much of you actually needs to stay the same for you to remain you?

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton