No scientific test for consciousness exists even in humans — how can we declare AI unconscious?

burny_tech · x · 2026-09-21

The author challenges how anyone can confidently answer "no" on AI consciousness: we lack a proper falsifiable scientific test for consciousness (phenomenal experience) even in humans, so any yes/no claim requires an established theory of consciousness to back it.

He adds that we don't fully understand what neural networks do — matrices trained by gradient descent exhibit emergent higher-order regularities and laws at other levels of abstraction, and both mechanistic interpretability and deep learning theory remain full of open problems. Assuming we know everything about these systems is oversimplification and overconfidence.

Original post →

More from AGI Musings

AGI Musings channel →