claim
active
claim:a-loss-minimizing-model-points-its-most-consequential-weights-in-good-directions-and-has-little-reason-to-police-weights-barely-reaching-outputA loss-minimizing model points its most consequential weights in good directions and has little reason to police weights barely reaching output
Interpretation of effectiveness tail result.
Source paper
extracted_fromNicholas L. Turner · Jeffrey Wu · Joshua Batson
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Three regimes by effectiveness.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Ethical implication about the nature of AI training experience if the thesis holds
- Extrapolation of scaling predictive models to AGI.
- Safety auditing caveat.
- Fig. 22.
- Basis of phenomenological argument.
- Fig. 23.
- Opening sentence defining self-evidencing.
- Do we assume that the model knows something we don't, or are these interference weights contributing noise?question0.748Motivating question.