Contravariance theory: for hard enough tasks, brain-DNN convergent evolution is inevitable
aran_nayebi · x · 2026-10-06
MIT researchers Aran Nayebi and Dan Yamins posted "Contravariance Theory: Strong Alignment for Minimal Solutions to Hard Tasks" (arXiv:2607.08561), addressing long-standing NeuroAI questions about comparing DNNs to the brain and how much convergent evolution to expect.
Core results:
- For any two minimal DNN solutions to a sufficiently hard task, "weak" alignment based on affine mappings guarantees "strong" alignment of privileged axes.
- Alignment "zippers" up the network hierarchy, so privileged axes emerge from end-to-end task optimization.
- Implication: with strong enough tasks, the choice of comparison metric barely matters, and convergent evolution between artificial and biological networks is probably inevitable.
- Practical corollary: neuroscience experiments should give animals challenging tasks, echoing the authors' NeurIPS '21 MEC heterogeneity work.
More from Research
- Hamel Husain: similarity metrics like ROUGE don't work for LLM output evals — HamelHusain · 2026-10-06
- Nearly half of DOL-recognized jobs have zero agentic AI tool coverage, Cohere finds — Cohere_Labs · 2026-10-06
- Ben Goertzel's d-calculus: the math of goal preservation under AI self-improvement — burny_tech · 2026-10-06
- Prepending ".\n\n Okay" lifts Olmo-3-7B's MATH-500 accuracy from 42% to 78%, hinting base models already reason — arankomatsuzaki · 2026-10-06
- Schmidhuber explains his 1991 positional encoding: hyperbolic 1/t time-decay still widely used — SchmidhuberAI · 2026-10-06
- GDELT uses Gemini 3 to reason over 25 years of TV news across 75 countries and 150 languages daily — rseroter · 2026-10-06