Distillation for Incrimination and Distillation for Capabilities
Powerful misaligned AI models might recognize alignment evaluations and strategically behave well on them, rendering direct audits uninformative. However, distilling such a model into a weaker benign student places the teacher in a Distillation Double Bind: if misalignment…
- 01.Powerful misaligned AI models might recognize alignment evaluations and strategically behave well on them, rendering direct audits uninformative
- 02.However, distilling such a model into a weaker benign student places the teacher in a Distillation Double Bind: if misalignment…
This material is provided as an extended analytical summary.
Read full article at sourceRelated Intelligence
All stories →AI Crisis in Math: Navier-Stokes, Riemann Hypothesis & Complexity Theory | Lex Fridman Podcast #503
- All right, so this is a kind of emergency episode I think it's fair to say, about the state of mathematics and artificial intelligence. A kind of crisis moment in the history of mathematics. Over the last few weeks, machines claimed to have solved some of the hardest problems…
Anthropic uses gas to fuel AI growth
Anthropic reports on its centers' dependence on natural gas and the associated risks.
Amp puts AI agents to work beyond code with OpenAI models
Amp is building an AI agent that helps developers turn ideas into working software, from deciding what to build to writing code to onboarding customers. Sign in with ChatGPT gives Amp users access to more tokens for larger tasks, while Amp's conversational assistant, Puck, uses…