concept
active
concept:richens-everitt-2024

Richens & Everitt (2024)

Showed robust agents learn causal world models; related to PRH claim about convergence to reality model

Neighborhood — ranked by edge-count

Related by similarity (7)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

  • Posited that generality of representations rather than particular task explains alignment with biological representations
  • Introduced the Contravariance principle; closely related theoretical foundation for multitask scaling hypothesis
  • Introduced CKA and observed that model alignment increases with model scale and dataset size
  • Showed word embeddings of visual concept names can be isometrically mapped to image embeddings, and developed framework for efficient concept extraction
  • Found Rosetta Neurons — neurons activated by same patterns across diverse vision models
  • Source of the HMM parameters for normal and chronic pain models used in this paper
  • Cited for analysis of AI and economic growth relevant to Malthusian dynamics of digital minds