Anthropomorphizing AI: Useful or Misleading? It Depends on the Level of Abstraction

sebkrier · x · 2026-09-11

Researcher sebkrier argues that debates over whether to anthropomorphize AI systems—"it's just matrix multiplications" vs. "it's more than that"—often fail because participants don't specify which phenomena or level of abstraction their claims apply to.

His position: anthropomorphic descriptions can be pragmatically useful for interaction but may obscure differences that matter for understanding how systems work, while purely mechanistic descriptions can wrongly dismiss functionally relevant similarities whose significance varies by context.

Original post →

More from AGI Musings

AGI Musings channel →