AI Snake Oil authors launch nonprofit STABL, targeting the "missing quadrant" of AI debate
AI Snake Oil · rss · 2026-11-11
The authors of the AI Snake Oil blog announce a new nonprofit research organization, STABL. They frame the AI debate on a 2x2 chart: whether you take AI's transformative potential seriously, and whether you take downstream bottlenecks to AI's impacts seriously.
- Top-left (technological determinism): superintelligence unilaterally obliterates bottlenecks, compressing a century of progress into a few years — a view shared by many AI CEOs and, ironically, by some of the industry's strongest critics who focus solely on slowing capabilities.
- Bottom-right: treats existing organizations, professions, and governance as fixed, insisting technology mold to them (e.g., resisting self-driving cars, banning AI legal advice).
- Top-right (STABL's stance): takes frontier AI seriously as transformative, while holding that institutional change and societal resilience — not raw capability — will bottleneck societal outcomes. They argue this quadrant is critically underserved, fueling polarization.
Using medicine as an example: determinists claim superintelligence will rapidly cure disease; skeptics correctly note the bottleneck is clinical trials, not drug discovery. STABL argues identifying bottlenecks is only the starting point — the real question is how to reform institutions (e.g., in-silico testing, regulatory redesign) so AI's benefits materialize.
Four research directions:
- Measuring progress toward recursive self-improvement (RSI): via CRUX, frequent open-ended evaluations of long, messy tasks. First results show frontier agents cannot yet conduct open-ended AI research — expert reviewers unambiguously rejected agent-written papers. They also call for comparing causal models of AI progress and studying how data bottlenecks shape recursive improvement.
- Empirical research on the pace of progress and diffusion.
- Analysis and advocacy for risk mitigation beyond alignment and non-proliferation, with resilience as the most neglected defense.
- Measuring downstream outcomes of AI adoption, studying labor-market disruption, and proposing ways to build resilience and revitalize institutions.
More from AGI Musings
- AI 2027 Team's Q3.5 Update: Intelligence Explosion Will Happen in 2027 After All — birchlse · 2026-11-11
- Investor Alsop: far less convinced now that AI will leave humans jobless — StewartalsopIII · 2026-11-11
- MATS fellow essay: four reasons UBI isn't a durable fix for AI-driven unemployment — luke_drago_ · 2026-11-11
- JPMorgan sees under 0.5% cumulative TFP gain from AI over next decade — GaryMarcus · 2026-11-11
- L Somerset proposes testable inside-view vs outside-view framework for agent interactions — SucceededMind · 2026-11-11
- Josh Siegle proposes an "ImageNet for cell types" to link neural activity to cell identity — patrickmineault · 2026-11-11