Sequoia Capital

Will AI Ever Actually Be Conscious? Demis Hassabis, Google DeepMind

Demis Hassabis explores whether artificial systems could ever be truly conscious, emphasizing that consciousness remains an ill‑defined philosophical concept. He outlines core components—self‑awareness, a sense of self versus other, and temporal continuity—as likely necessary but not sufficient conditions. Hassabis distinguishes between behavioral mimicry and genuine experience, noting that AI may soon exhibit actions indistinguishable from conscious beings, yet it lacks the biological substrate that underpins human phenomenology. He argues that behavior alone cannot confirm consciousness, echoing long‑standing philosophical debates. Referencing conversations with philosophers, including the late Daniel, Hassabis points out that we assume others are conscious because they behave similarly and share our substrate. This substrate argument creates a barrier to accepting artificial consciousness, as machines operate on fundamentally different hardware. The discussion underscores the challenge of creating metrics for experiential consciousness in post‑AGI systems and highlights the broader implications for AI safety, ethics, and regulation. Without a clear definition, claims of machine consciousness remain speculative, urging the field to develop new theoretical and empirical tools.