AI Resilience
airesilience.net · 1,640 words · saved by 1 readers
Building resilience for the age of AI
Our vision for Epistemic resilience has (at least) two dimensions. One is advancing and protecting the epistemic ceiling: accelerating scientific progress while ensuring that the discoveries of vastly superhuman, automated intelligences are trustworthy and secure. Another is advancing and protecting epistemic diffusion: increasing the extent to which the whole of society is able to to discover, track, understand, share, and defend information and lines of reasoning that make it possible to reliably navigate the future. This civilization has done both. At the frontier, AI-powered science…
saved by
related reading
- AI Tools for Existential Securityforethought.org
- AI and Epistemics: The Good, Bad and Uglyforethought.org
- Where I agree and disagree with Eliezer — LessWronglesswrong.com
- BEWA: A Bayesian Epistemology-Weighted Artificial Intelligence Framework for Scientific Inferencearxiv.org
- Field-building and the epistemic culture of AI safetyfirstmonday.org
- Lab Leaks, Black Holes, and Eggs: Epistemic Case Study Competition — LessWronglesswrong.com
- AI Tools for Trust: Community Notes, Rhetoric Detection & Moreforethought.org
- Safety as a Scientific Pursuit - by Tom McGrathbanburismus.substack.com
- Preparing for Launch | IFPifp.org
- AI Robustnessdeepgram.com
- RESI Overviewresi.org
- [2004.07213] Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claimsarxiv.org