Would Sensory Capabilities Change Model Reasoning?

oyhsu · x · 2026-07-16

This post discusses a classic cognitive question: if a model could "smell," would its reasoning improve? The author believes this question actually reveals how one fundamentally understands "cognition" itself.

They offer a more specific but equally analogous scenario: if a model could "touch," would it be better at math? Overall, the thought experiment uses sensory expansion to reverse-engineer our intuitions about the mechanisms underlying model reasoning.

Original post →

More from AGI Musings

AGI Musings channel →