Can AI Ever Be Conscious? | Lucia Melloni
Why It Matters
The analysis frames AI consciousness as a testable scientific question with real design and policy implications: if certain architectures could produce subjective experience, developers and regulators may need new ethical and safety frameworks. Moreover, misattributing consciousness to current systems risks both misplaced trust and regulatory missteps.
Summary
Neuroscientist Lucia Melloni says distinguishing intelligence from sentience is essential when assessing whether AI could be conscious. Researchers can apply existing theories of consciousness—such as global workspace, integrated information theory (IIT), predictive processing and higher-order thought models—to search for indicators or architectures associated with subjective experience. Current AI lacks the neural-like architectures and integrated information levels these theories predict would be necessary, so present systems are unlikely to be conscious; however, Melloni warns future multi-agent or differently designed systems could meet those criteria. She also cautions that anthropomorphism and philosophical uncertainty about other minds make definitive attribution of consciousness inherently difficult.
Comments
Want to join the conversation?
Loading comments...