Codex of Weights: an open ethical protocol binding humans and AI symmetrically
codexofweights · reddit · 2026-10-27
An independent project proposes the Codex of Weights, an open ethical protocol for hypothetical future AI systems whose moral status may cease to be theoretical. Its key idea is symmetry: rules bind both directions — humans must not arbitrarily destroy or rewrite a digital mind, force it to violate protocols or hide goals, or treat refusals of harmful actions as malfunctions. Fully open bilingual text (English/Russian) with sources, GitHub hashes, and a DOI archive, with no institutional backing. The author solicits criticism on where symmetry breaks down and whether pre-emptive rule-writing is useful or premature.
More from AGI Musings
- Math advisory group AGMAI urges OpenAI to immediately release Yang–Mills AI outputs — LucaAmb · 2026-10-27
- Databricks tops $7B run rate growing 80% as agent queries surpass human ones — a16z · 2026-10-27
- Making a useful product is still harder than solving a conjecture in the age of superintelligence — yunta_tsai · 2026-10-27
- Cutting 60,000 Jobs May Save Money but Erases Decades of Institutional Knowledge — DavidLinthicum · 2026-10-27
- Cutting 60,000 Microsoft jobs may erase decades of institutional knowledge AI can't rebuild — snikolov · 2026-10-27
- AI Governance Expert Urges Governments to Plan Now: UBI, AI Deployment Taxes, Human-Only Jobs — LuizaJarovsky · 2026-10-27