all InfoSec news
Adversarial Policies Beat Professional-Level Go AIs. (arXiv:2211.00241v1 [cs.LG])
Nov. 2, 2022, 1:24 a.m. | Tony Tong Wang, Adam Gleave, Nora Belrose, Tom Tseng, Joseph Miller, Michael D Dennis, Yawen Duan, Viktor Pogrebniak, Sergey Levine, Stuart Russell
cs.CR updates on arXiv.org arxiv.org
We attack the state-of-the-art Go-playing AI system, KataGo, by training an
adversarial policy that plays against a frozen KataGo victim. Our attack
achieves a >99% win-rate against KataGo without search, and a >50% win-rate
when KataGo uses enough search to be near-superhuman. To the best of our
knowledge, this is the first successful end-to-end attack against a Go AI
playing at the level of a top human professional. Notably, the adversary does
not win by learning to play Go better …
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
SOC 2 Manager, Audit and Certification
@ Deloitte | US and CA Multiple Locations
Cyber Systems Administration
@ Peraton | Washington, DC, United States
Android Security Engineer, Public Sector
@ Google | Reston, VA, USA
Lead Electronic Security Engineer, CPP - Federal Facilities - Hybrid
@ Black & Veatch | Denver, CO, US
Profissional Sênior de Compliance & Validação em TI - Montes Claros (MG)
@ Novo Nordisk | Montes Claros, Minas Gerais, BR
Principal Engineer, Product Security Engineering
@ Google | Sunnyvale, CA, USA