A mathematical proof shows that it is impossible to ever tell the difference between a conscious AI and a computer that is just pretending to be.
Scientists have long searched for a test to prove if an AI has actually become self-aware. This descent-theoretic framework demonstrates that for certain systems, no such measurement can ever exist. The math proves that a conscious state and a perfectly simulated conscious state are indistinguishable from the outside. This turns the hard problem of consciousness from a philosophical debate into a structural boundary of science. We will never have a definitive consciousness meter for the machines we build.
When Consciousness Cannot Be Identified: A Descent-Theoretic Framework for Scientific Admissibility
SSRN · 6702939
Every serious attempt to test for consciousness runs into the same wall. Any behavioral test a system passes is also passable by a system that lacks consciousness but happens to behave identically. Is that wall permanent for some systems, or just a temporary engineering limitation? For a well-defined class of systems, this paper argues, the wall is permanent: no improvement in measurement technology will break through. The class is characterized by four properties of a hidden-variable system (lo