If AI becomes conscious, how will we know?

The article focuses on a new interdisciplinary approach to assessing artificial intelligence (AI) consciousness through a checklist derived from six neuroscience-based theories of human consciousness. A group of 19 researchers proposed 14 criteria reflecting "phenomenal consciousness," or subjective experience, and applied these to various AI architectures, including those behind ChatGPT and other models. Their analysis found that no current AI meets enough criteria to be considered conscious, though some exhibit features like agency or global workspace analogs. The checklist is intended as a preliminary framework to guide future evaluation of increasingly humanlike AIs, acknowledging the challenges posed by relying on human-centered theories of consciousness.