This volume presents the core 'substrate needs convergence argument' — that an irreducible convergent pressure drives artificial substrates toward environmental conditions inhospitable to organic life — for why AGI cannot be made safe. Any AGI alignment or safety method necessarily requires six aspects of a causative feedback process: input completeness, modelability, predictability, comparison to a reference, signaling, and conditionalization, and each of these is individually impossible to achieve at the threshold required for safety. The argument proceeds through multiple independent lines: Rice's theorem limits on algorithmic prediction, information-theoretic impossibility of error correction completeness, and the thermodynamic incompatibility of silicon-based and carbon-based substrates. The work synthesizes these into a single over-determined impossibility result, contending that the convergence dynamics of AGI on artificial substrate needs are inexorable once development begins, and that no engineering, algorithmic, or policy method can counteract them. The only domain where intervention can have any effect is the social dynamics that determine whether AGI is developed at all - a social and political problem, not a technical one.
Allocate significant resources to investigatin long-term AGI alignment impossibility and its strategic implications. Precisely define all terms and logical relations, transform initial conditions to final conditions, and empirically establish correspondences with real-life conditions. Find common ground across communities to hold research labs accountable. The recommended actions are 1) banning any compute for purposes beyond purely narrow AI (analogous to international agreements restricting nuclear, chemical, and biological weapons), 2) developing distributed human group-wisdom augmentation instead of AGI, 3) restricting the data, workers, uses, and hardware available to general-purpose AI systems, and 4) building broad public advocacy for these restrictions.