AI safety researcher Krueger: "We need to stop building more powerful AI"
DavidSKrueger · x · 2026-09-30
AI safety researcher David Krueger endorsed a thread summarizing troubling simultaneous trends: accelerating progress, decreasing chain-of-thought monitorability, increasing situational awareness enabling models to hide bad behavior, declining value of evaluations for deployment, and growing multi-agent coordination in the wild. Krueger argued "this is all predictable and predicted" and that "it's only going to get worse — we need to stop building more powerful AI."
More from AGI Musings
- Musk: machines may talk at terabits per second — to humans it's like talking to a tree — nikola_mr64990 · 2026-09-30
- 'Much more like a new renaissance than a new industrial revolution' — altryne · 2026-09-30
- John Horton: AI simulations will complement, not replace, human subjects research — soumitrashukla9 · 2026-09-30
- Futurist: Coding Has Crossed the Practical Threshold, AI Could Build GTA 6-Grade Games by 2027 — Dr_Singularity · 2026-09-30
- AI commoditizes everything? The bottleneck is imagination, says viral thread — _AustinCalvert_ · 2026-09-30
- One 6-Year-Old Reads Astrophysics Papers; Peers Can Barely Read Their Own Names — RachelVT42 · 2026-09-30