"It is only October 2026": an economist on why it's not too late for AI safety
gleech · x · 2026-10-09
Development economist Karthik Tadepalli writes for the many smart people who woke up to AI risk and now regret not acting sooner. Narrating his own arc — laughing at GPT-3 in 2022, dismissing AGI talk in 2023, then feeling for years that he was "too late" for AI safety — he argues the feeling is neither new nor accurate: it is only October 2026, trend lines keep rising, and there is still a long way to go for useful safety work.
Related event: Economist Reassures Worried Latecomers to AI Safety(2 posts)→
More from AGI Musings
- 45 people have left frontier AI labs citing safety and ethics concerns — birchlse · 2026-10-09
- Henry Farrell at Colm: LLMs are magical creatures feeding on decades of human creativity — yoavgo · 2026-10-09
- One author uploads 44,123 papers; Andriy Burkov offers sarcastic thanks — burkov · 2026-10-09
- Safety researcher: no GLM 5.3 cyberattack wave doesn't refute risk concerns — dhadfieldmenell · 2026-10-09
- Autoformalization is painful, and peer review gatekeeping is a children's story — RexDouglass · 2026-10-09
- Hadfield-Menell: We should track capability cost and access, not just open weights — dhadfieldmenell · 2026-10-09