Irving, 2018 — {AI} Safety via Debate

DOI: <a href="https://doi.org/10.48550/arXiv.1805.00899" target="_blank" rel="noopener noreferrer">https://doi.org/10.48550/arXiv.1805.00899
URL: <a href="https://arxiv.org/abs/1805.00899" target="_blank" rel="noopener noreferrer">https://arxiv.org/abs/1805.00899

Presents a scalable oversight proposal or failure mode for supervising systems stronger than their overseers.

Publication links

Presents a scalable oversight proposal or failure mode for supervising systems stronger than their overseers.