
Sign up to save your podcasts
Or


The paper investigates whether current safety training techniques can detect and remove deceptive behavior in AI systems. The study finds that backdoored behavior can persist in large language models, even with standard safety training techniques, and adversarial training can hide the unsafe behavior.
https://arxiv.org/abs//2401.05566
YouTube: https://www.youtube.com/@ArxivPapers
TikTok: https://www.tiktok.com/@arxiv_papers
Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016
Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
By Igor Melnyk5
33 ratings
The paper investigates whether current safety training techniques can detect and remove deceptive behavior in AI systems. The study finds that backdoored behavior can persist in large language models, even with standard safety training techniques, and adversarial training can hide the unsafe behavior.
https://arxiv.org/abs//2401.05566
YouTube: https://www.youtube.com/@ArxivPapers
TikTok: https://www.tiktok.com/@arxiv_papers
Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016
Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers

977 Listeners

1,993 Listeners

443 Listeners

113,121 Listeners

10,254 Listeners

5,576 Listeners

221 Listeners

51 Listeners

101 Listeners

475 Listeners