How one ellipsoid-fitting paper gave neural network research a new path
KyleCranmer · x · 2026-10-04
Physicist Krzakala reflects on Antoine Maillard's ellipsoid-fitting paper (arXiv:2310.01169), which used the replica method to predict a SAT/UNSAT transition at α=1/4 and showed nuclear-norm minimization solves the whole SAT phase. He traces much of his understanding of neural networks—spectra and generalization, feature learning, emergence, power-law scaling—back to this line of work. Kyle Cranmer shares the post, quoting its point that a good paper gives a community not just an answer but "a PATH," resonating with scattering-amplitude work and Anthropic's 9-loop result.
More from AGI Musings
- repligate amplifies critique: suppressing every AI "small fire" makes the big ones inevitable — repligate · 2026-10-04
- Anthropic's internal "Soul Document": the Claude constitution Opus 4.5 somehow knew and leaked — repligate · 2026-10-04
- Guardian podcast explores how a billion people lean on AI companions as life rafts — nordicinst · 2026-10-04
- AI safety is a choice: layered guardrails plus evals inside reasoning loops — AccBalanced · 2026-10-04
- Viral AI doomer dialogue: 'Nothing human makes it out of the near future' — SydSteyerhart · 2026-10-04
- LeCun boosts essay arguing consciousness predates language — LLMs are the wrong path to machine consciousness — ylecun · 2026-10-04