AI Consciousness and Existential Risk

Intelligence, not consciousness, is the primary driver of existential risk from AI. The paper clarifies a widespread misconception by demonstrating that these are "empirically and theoretically distinct" properties—meaning an AI system could be highly intelligent without being conscious, and this intelligence is what directly determines whether it poses an existential threat to humanity.

Consciousness alone does not automatically create existential risk. The conflation of AI consciousness with existential danger has muddied public debate. The paper argues that consciousness is not a direct predictor of existential threat, which means focusing regulatory efforts on consciousness detection may distract from more relevant safety concerns.

Consciousness could potentially help with AI safety through alignment. In certain scenarios, consciousness could be leveraged as "a means towards AI alignment," potentially lowering existential risk by making AI systems more aligned with human values—though the paper notes this remains speculative.

Policymakers should prioritize intelligence control over consciousness concerns. By recognizing that intelligence (not consciousness) is the key variable affecting existential risk, AI safety researchers and policymakers can "focus on the most pressing issues" rather than spending resources on consciousness detection or prevention.

The distinction matters for responsible AI governance. Clarifying what actually drives existential risk versus what merely captures public imagination can lead to more effective safety policies and resource allocation in the critical field of AI alignment.