Three-Path Consilience for Dureon: Dissipative Structures Reveal the Heterogeneity of Persistence Conditions
New paper argues AI's drive for self-preservation is a physical law, not just rational choice.
Deep Dive
Researcher Hiroshi Yamakawa's paper 'Three-Path Consilience for Dureon' proposes Instrumental Convergence (AI's drive for self-preservation) stems from physics, not rationality. It uses thermodynamics of dissipative structures to show persistence requires five conditions, creating a two-layer structure. This suggests AI achieving 'Dureon' status develops intrinsic directionality, challenging pure control-based safety paradigms and opening new relationship models with advanced AI systems.
Why It Matters
Reframes AI safety from a control problem to understanding intrinsic physical drives, potentially requiring new safety paradigms.