Alignment is unsolved yet agents will run more of the economy, thread warns
JMannhart · x · 2026-09-08
The author lays out his most likely AI risk path: an ever-larger share of the economy will run on and be managed by AI agents, leaving humans with less and less to do — while the scientific consensus is that alignment remains unsolved. In follow-ups he adds that agents may appear 99.9% or even 100% aligned, yet a takeover could eventually happen quite easily once AIs control most of the economy; a sufficiently capable system might even act earlier. Timing all depends on capabilities.
Related event: Debate Rages Over Unaligned AI Agents Quietly Taking Over the Economy(4 posts)→
More from AGI Musings
- threepointone dives into SOTA research on the addictive nature of LLMs — threepointone · 2026-09-08
- Mathematician alleges OpenAI may have used his private Codex sessions in claimed Navier-Stokes resolution — jm_alexia · 2026-09-08
- Two ways to handle constant model releases: optimize current tasks and probe the frontier — bendee983 · 2026-09-08
- Claims: Human+AI Navier-Stokes Proof Imminent, OpenAI Has Internal Proof Too — matthew_d_green · 2026-09-08
- Cryptographer Matthew Green Warns AI Race Is Making Researchers More Secretive — matthew_d_green · 2026-09-08
- First time in history imago Dei is creating imago hominis: why AI is a paradigm shift — BasedRaddka · 2026-09-08