673c751b76
## Summary Comprehensive audit of all 86 foundation claims across 4 subdomains. **Changes:** - 7 claims moved (3 → domains/ai-alignment/, 3 → core/teleohumanity/, 1 → domains/health/) - 4 claims deleted (1 duplicate, 3 condensed into stronger claims) - 3 condensations: cognitive limits 3→2, Christensen 4→2 - 10 confidence demotions (proven→likely for interpretive framings) - 23 type fixes (framework/insight/pattern → claim per schema) - 1 centaur rewrite (unconditional → conditional on role complementarity) - All broken wiki links fixed across repo **Review:** All 4 domain agents approved (Rio, Clay, Vida, Theseus). Pentagon-Agent: Leo <76FB9BCA-CC16-4479-B3E5-25A3769B3D7E>
34 lines
No EOL
4.7 KiB
Markdown
34 lines
No EOL
4.7 KiB
Markdown
---
|
|
description: Current alignment approaches are all single-model focused while the hardest problems preference diversity scalable oversight and value evolution are inherently collective
|
|
type: claim
|
|
domain: ai-alignment
|
|
created: 2026-02-17
|
|
source: "Survey of alignment research landscape 2025-2026"
|
|
confidence: likely
|
|
---
|
|
|
|
# no research group is building alignment through collective intelligence infrastructure despite the field converging on problems that require it
|
|
|
|
The most striking gap in the alignment landscape as of 2025-2026: virtually no one is building alignment through collective intelligence infrastructure. The closest attempts are partial. Since [[democratic alignment assemblies produce constitutions as effective as expert-designed ones while better representing diverse populations]], CIP has demonstrated that democratic input works mechanically -- but this remains one-shot constitution-setting, not continuous architecture. Since [[community-centred norm elicitation surfaces alignment targets materially different from developer-specified rules]], STELA has shown that inclusive deliberation produces different outputs -- but it does not build the infrastructure for ongoing participation. Polis does consensus-mapping through statement submission and voting. Some multi-agent debate frameworks exist under the scalable oversight umbrella. The Cooperative AI Foundation studies multi-agent coordination. But none of these constitute a distributed architecture where alignment emerges from collective participation.
|
|
|
|
What does not exist: no system where contributor diversity structurally prevents value capture; no implementation of continuous value-weaving at scale; no infrastructure for collective oversight of superhuman AI components; no architecture where alignment is a property of the coordination protocol rather than a property trained into individual models. Since [[universal alignment is mathematically impossible because Arrows impossibility theorem applies to aggregating diverse human preferences into a single coherent objective]], the impossibility of aggregation makes collective infrastructure -- which preserves diversity rather than aggregating it -- the only viable path.
|
|
|
|
This gap is remarkable because the field's own findings point toward collective approaches. Since [[RLHF and DPO both fail at preference diversity because they assume a single reward function can capture context-dependent human values]], diverse preference representation is needed. Since [[scalable oversight degrades rapidly as capability gaps grow with debate achieving only 50 percent success at moderate gaps]], distributed oversight is needed. Since [[the alignment tax creates a structural race to the bottom because safety training costs capability and rational competitors skip it]], structural alignment is needed to eliminate the tax.
|
|
|
|
The alignment field has converged on a problem they cannot solve with their current paradigm (single-model alignment), and the alternative paradigm (collective alignment through distributed architecture) has barely been explored. This is the opening for the TeleoHumanity thesis -- not as philosophical speculation but as practical infrastructure that addresses problems the alignment community has identified but cannot solve within their current framework.
|
|
|
|
---
|
|
|
|
Relevant Notes:
|
|
- [[AI alignment is a coordination problem not a technical problem]] -- the gap in collective alignment validates the coordination framing
|
|
- [[collective superintelligence is the alternative to monolithic AI controlled by a few]] -- the only project proposing the infrastructure nobody else is building
|
|
- [[RLHF and DPO both fail at preference diversity because they assume a single reward function can capture context-dependent human values]] -- collective approaches address this specific failure
|
|
- [[the alignment tax creates a structural race to the bottom because safety training costs capability and rational competitors skip it]] -- structural alignment eliminates the tax
|
|
- [[democratic alignment assemblies produce constitutions as effective as expert-designed ones while better representing diverse populations]] -- the closest existing work, but still one-shot not continuous
|
|
- [[community-centred norm elicitation surfaces alignment targets materially different from developer-specified rules]] -- demonstrates what inclusive infrastructure reveals, but does not build the infrastructure
|
|
- [[universal alignment is mathematically impossible because Arrows impossibility theorem applies to aggregating diverse human preferences into a single coherent objective]] -- the impossibility of aggregation makes collective infrastructure the only viable path
|
|
|
|
Topics:
|
|
- [[livingip overview]]
|
|
- [[coordination mechanisms]]
|
|
- [[domains/ai-alignment/_map]] |