carado.moe

tammy's blog about
AI alignment, utopia, anthropics, and more;

also on twitter, lesswrong, rss, and at

the QACI alignment plan: table of contents

this post aims to keep track of posts relating to the question-answer counterfactual interval proposal for AI alignment, abbreviated "QACI" and pronounced "quashy". i'll keep it updated to reflect the state of the research.

this research is primarily published on the Orthogonal website and discussed on the Orthogonal discord.

as a top-level view of QACI, you might want to start with:

• an Evangelion dialogue explaining QACI
• a narrative explanation of QACI
• Orthogonal's Formal-Goal Alignment theory of change
• formalizing the QACI formal-goal

the set of all posts relevant to QACI includes:

• as overviews of QACI and how it's going:
  • state of my research agenda
  • problems for formal alignment
  • the Formal-Goal Alignment theory of change
  • the original post introducing QACI
• on the formal alignment perspective within which it fits:
  • formal alignment: what it is, and some proposals
  • clarifying formal alignment implementation
  • on being only polynomial capabilities away from alignment
• on the blob location problem:
  • QACI blobs and interval illustrated
  • counterfactual computations in world models
  • QACI: the problem of blob location, causality, and counterfactuals
  • QACI blob location: no causality & answer signature
  • QACI blob location: an issue with firstness
• on QACI as an implementation of long reflection / CEV:
  • CEV can be coherent enough
  • some thoughts about terminal alignment
• on formalizing the QACI formal goal:
  • a rough sketch of formal aligned AI using QACI with some actual math
  • one-shot AI, delegating embedded agency and decision theory, and one-shot QACI
• on how a formally aligned AI would actually run over time:
  • AI alignment curves
  • before the sharp left turn: what wins first?
• on the metaethics grounding QACI:
  • surprise! you want what you want
  • outer alignment: two failure modes and past-user satisfaction
  • your terminal values are complex and not objective
• on my view of the AI alignment research field within which i'm doing formal alignment:
  • my current outlook on AI risk mitigation
  • a casual intro to AI doom and alignment