AGI Discussions on Data and Defensive Capabilities
RyanGreenblatt · x · 2026-07-13
This in-depth post continues the discussion on AI training data, evaluation data, and whether "improving defensive capabilities is sufficient."
Core viewpoints:
- Previously, the general consensus was to oversample certain data while filtering out other data (like evals); today's views are more nuanced but directionally similar.
- The author argues that by the time many issues become "visible," it is often too late or requires drastic intervention.
- Merely improving defensive capabilities in a broad sense is usually insufficient to truly solve such problems, though it still helps to some extent.
Related event: AI Community's Shift in Data Strategy(3 posts)→
More from AGI Musings
- Benjamin Todd: AI progress beat superforecasters over 18 months — ben_j_todd · 2026-10-26
- AI researchers guessed robots wouldn't outrun humans until 2034 — it already happened — ben_j_todd · 2026-10-26
- AI folks, stop saying "biology" when you mean drug development — KevinKaichuang · 2026-10-26
- ESR on the Bugpocalypse: locking down frontier models only hands the game to Chinese open weights — colinmcnamara · 2026-10-26
- Anthropic: junior tech hires fell from 27% to 9% as agent hours hit 6x employee hours — thione · 2026-10-26
- Anthropic's Jascha: you may have only 1–5 years of intellectual contribution before AGI — GabGarrett · 2026-10-26