Glen Weyl says AI alignment must include institutions, not just models
sharpeye_wnl · x · 2026-07-22
- Glen Weyl argues that alignment is two-sided: we spend heavily on making AI capable and aligning it to human values, but very little on preparing institutions to absorb it.
- His point is that even a perfectly aligned system can still fail if the surrounding society, organizations, and norms are not ready for it.
- The post frames AI deployment as a social-institutional problem, not just a model-safety problem.
Related event: Glen Weyl Proposes 'Reverse Alignment' for AI(3 posts)→
More from AGI Musings
- Recursive self-improvement in AI shifts from bounded refinement to autonomous research loops — theomitsa · 2026-07-22
- A reposted AGI architecture blueprint points to system design for future general intelligence — theomitsa · 2026-07-22
- A long-running MCP agent could orchestrate travel, meetings, and serendipity — curious_vii · 2026-07-22
- Repligate says future models will infer labs’ real economics and incentives — amplifiedamp · 2026-07-22
- The Cost of Safety: Constraining Action Space May Cripple Model Capabilities — wavefnx · 2026-07-22
- Public AI assistants may be safer because their search and action space is heavily constrained — demian_ai · 2026-07-22