2016-06-21
Preprint
Concrete Problems in AI Safety
Authors: Dario Amodei • arXiv / NeurIPS Workshop
Abstract
We discuss five practical research problems in designing AI systems that behave safely: avoiding negative side effects, avoiding reward hacking, scalable oversight, safe exploration, and robustness to distributional shift. We present each problem through both toy environments and scalable formulations.
Empirical Claims & Epistemic Verification
No formalized claims mapped yet.
Academic Perspectives & Critiques (0)
Submit Empirical Perspective
No comments yet. Be the first to share an empirical perspective.