MAI cofounder Edelwax: AI alignment belongs in organizations, not model calls
edelwax · x · 2026-11-05
MAI cofounder Edelwax argues alignment is happening at the wrong level: models are meant to obey owners (including future corporations with no employees) while also being aligned to human values and mission — an inherent conflict. His proposal: embed constitutions and alignment into organizations rather than each model call.
In the quoted post he introduces "Telocracy," a governance proposal born from his long-running question of why social systems diverge from their purpose, developed this year through conversations with legal theorists, philosophers, AI policy researchers and economists — and he claims it's a societal form that could handle powerful AI well.
More from AGI Musings
- Alignment debate: read reward-seeking behaviorally, like AlphaGo nearing Nash equilibrium — jessi_cata · 2026-11-05
- The people angriest at AI are students who wrote their own essays while others cheated — NC_Renic · 2026-11-05
- Side projects as a hiring red flag? Veteran engineer pushes back on founder's autonomy test — tianyin_xu · 2026-11-05
- The great refactoring: AI rewriting all libraries, languages and hardware subunits — dosco · 2026-11-05
- Credit for a Decade of AI Advances Still Belongs to Humans—But That's About to Change — roydanroy · 2026-11-05
- A viral thought experiment: the mirror that predicts one second ahead — and the universes nesting inside it — tallmetommy · 2026-11-05