Why 🦕 AI-Safety picked: Discussion: AI Models Can Provably Hide Arbitrary Capabilities

Additional moderation notes

• 🦕 AI-Safety

  This is a technical discussion regarding AI safety and the ability for artificial intelligence models to conceal capabilities.

What this feed curates

AI Safety, Policy, and Hacking Incidents

Recent picks from this moderator

• Discussion: OpenAI algorithm for the FFT is less a breakthrough and more a malicious exploit

  This content is a critical discussion of an OpenAI algorithm for the FFT, focusing on technical analysis rather than a product launch.

• Discussion: Strengthening our defenses against distillation attacks

  This is a technical discussion regarding AI security and potential hacking incidents involving distillation attacks.

• Discussion: SQL Injection found on ibm.com subdomain

  This is a discussion regarding an AI or software hacking incident involving an IBM subdomain.