Researcher: as model risks escalate, we still have tools we trust today
sleepinyourhat · x · 2026-09-23
A researcher notes that while the field worries about its ability to keep up with the escalating risks that come with more capable models, there remain many tools we trust at current capability levels.
More from AGI Musings
- Builders break ground on a 'post-AGI city' in Uruguay, slated to start in 2027 — abhiadesai · 2026-09-23
- Dietterich: Leveson's dynamic safety is essentially alignment — tdietterich · 2026-09-23
- Don't outsource your AI strategy to vendors that sell tokens — springrod · 2026-09-23
- Should human-only discussion spaces exist? Debate over AI quietly infiltrating discourse — gerardsans · 2026-09-23
- Tom Dietterich defines safe systems: harms must stay below socially acceptable levels despite disturbances — tdietterich · 2026-09-23
- Mathematics is getting automated away faster than B2B SaaS, and nobody saw it coming — teodorio · 2026-09-23