The Inside View cover image

[Crosspost] Adam Gleave on Vulnerabilities in GPT-4 APIs (+ extra Nathan Labenz interview)

The Inside View

00:00

Exploiting Vulnerabilities in Superhuman Go Playing AIs

This chapter discusses vulnerabilities in superhuman go playing AIs, highlighting the concept of gray box access and how adversaries can exploit weaknesses to defeat the AI. It explores the challenges of securing AI systems like Go engines and the evolving nature of attacks that target these systems. The conversation also touches on adversarial attacks on open source models, the trade-offs in achieving superhuman capability and robustness, and the importance of addressing safety issues in AI design.

Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner
 Get the app