RichardMCNgo: Why you don't need to stay at AGI labs for safety
nitarshan · x · 2026-08-27
RichardMCNgo argues that AI safety researchers don't need to stay at top AGI labs like Anthropic, OpenAI, or DeepMind to do good work, gain influence, issue warnings, or implement guardrails. Key points include: scaling research can be done on public models; high-pressure environments kill creativity; warnings will soon be abundant; racing to beat 'bad guys' is part of the problem; and many are eager to replace those implementing short-term mitigations.
More from AGI Musings
- RyanGreenblatt: Lack of Tools to Oversee AI Swarms — RyanGreenblatt · 2026-08-27
- Top surgeon predicts AI will be mandatory in care within 3 years — mattbeane · 2026-08-27
- Opinion: University interns outperform AI Agents — gaganghotra_ · 2026-08-27
- Local Agents as Gateway to Frontier Models — AravSrinivas · 2026-08-27
- Reversion to Human-ness as AI Blur Truth — zacharynado · 2026-08-27
- NYT Opinion: The AI Debt Binge Is Endangering the Economy — GaryMarcus · 2026-08-27