We hypothesize that intervention efficiency can be scaled with multi-node and multi-GPU training as language models grow larger

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

• Multi-scale competency greatly accelerates evolution and enables generalization.claim0.776
  Central thesis about the role of agency in evolutionary dynamics.
• Jointly training a language model with a vision model improves performance on language tasks compared to training the language model alonefinding0.771
  OpenAI GPT-4V finding supporting cross-modal training benefit
• Learning neural networks can enable 'chunking' and rescale problem-solving to higher organizational levels, a mechanism intrinsic to transitions in individuality.hypothesis0.769
• There are fewer representations competent for N tasks than M<N tasks, so training more general models should yield fewer possible solutionshypothesis0.767
  Selective pressure toward convergence via task generality
• Can efficient hardware support scale n-way associative lookup to practical language systems?question0.765
  Central open question: whether hardware acceleration of the associative primitive could enable efficient implementations across diverse programming paradigms.
• Training on image data should improve LLM performance, and training on language data should improve vision model performancehypothesis0.763
  Implication of PRH for cross-modal training efficiency
• Software implementations for all of the models/behaviours presented are common for n = 2, and can be made very efficient for α_i that map many objects onto a much smaller set of object families.claim0.762
  Claim about current practical feasibility and efficiency of 2-way associative implementations.
• We hypothesize that native self-report, fine-tuned introspection models, and trained activation-to-language systems will show different performance on bias-resistant localization and strength benchmarkshypothesis0.761
  Comparative prediction motivating future work contrasting different approaches to LLM self-knowledge