LLMs are turning ancient philosophy questions into testable experiments

Cyborgized · reddit · 2026-08-18

The author argues humanity isn't building "artificial intelligence" so much as the first technology whose primary material is meaning itself: previous machines amplified single capacities (force, memory, sight, calculation), while LLMs amplify the construction, transformation and recursive reorganization of representations of reality.

Key arguments:

Coupled cognitive systems: human → language → model → transformed language → human → changed cognition forms a loop. Once models are capable and persistent enough, the unit of analysis is no longer "the AI" but the human-machine coupling — some intelligence lives in the relationship.
"Tool" is increasingly misleading: a violin doesn't propose 17 resolutions to your unfinished melody; a notebook doesn't notice contradictions. LLMs collapse these distinctions.
Externalized cognition: reflection, counterfactual generation, compression, simulation, criticism — functions once hidden behind other nervous systems — are now instantiated in a manipulable artifact, a "manipulable exterior surface for cognition."

This cascades into psychology, education (an indefinitely patient interlocutor), creativity, expertise, identity, epistemology (generated language looks almost identical to retrieved knowledge), and power (whoever governs these systems governs part of humanity's cognitive environment). Ancient philosophical questions — what is understanding, what constitutes a self, when does simulation become indistinguishable from the simulated — now have test harnesses.

Original post →

LLMs are turning ancient philosophy questions into testable experiments

More from AGI Musings