Neural Atlas · About

An anatomical map of what we have built.

Neural Atlas maps the state of the art in artificial intelligence onto the human brain. You start from a cognitive function and arrive at the technology that tries to replicate it — never the other way around.

Concept

Most AI landscape maps are organised by company, model or industry vertical. They answer the question what exists? but not what does it do, biologically?. Neural Atlas inverts the perspective: every region of the brain corresponds to a cognitive or sensorimotor function, and each function is paired with the AI architecture that comes closest to performing it, annotated with a maturity verdict.

V1.3 covers eight regions — vision, language, reasoning, memory, motor control, emotional regulation, multimodal integration and procedural learning. They are the most developed branches of modern AI, and the ones where the comparison with the brain says the most. Future versions extend the mapping to the rest of the central nervous system, the autonomic system and the body.

Methodology

Each region has a structured profile: biological function, AI counterpart, current status, key milestones, concrete examples and remaining gaps. We assign a maturity level on a six-point scale:

  • Absent no known AI counterpart exists.
  • Embryonic research prototypes only.
  • Primitive production exists but is qualitatively far from biology.
  • Developing capability is improving fast on canonical benchmarks.
  • Mature competitive with or exceeding humans on most narrow tasks.
  • Superhuman measurably above the best humans.

The verdicts are editorial. They are anchored to the Stanford AI Index 2026 and the International AI Safety Report 2026 where possible, but the framing — what counts as "mature", where the hard part really is — is an interpretation. We argue our opinions; we do not claim consensus.

Sources

  • Stanford HAI — AI Index 2026 primary source for capability benchmarks, technical performance and adoption trends.
  • International AI Safety Report 2026 used for risk, alignment and oversight context.
  • Anthropic, OpenAI, Google DeepMind, Meta AI and Physical Intelligence release notes for Q1–Q2 2026 — used to verify concrete model and product milestones.
  • BrainFacts.org and the Society for Neuroscience — anatomical and functional definitions for the eight V1.3 regions.

Attribution

The 3D brain mesh is derived from Z-Anatomy (CC-BY-SA 4.0) by Gauthier Kervyn, Marcin Zielinski and Lluis Vinent — itself based on BodyParts3D (NBDC/RIKEN, CC-BY-SA 2.1 JP). The exported mesh is redistributed under CC-BY-SA 4.0 in compliance with copyleft.

Source code: MIT. Editorial content: CC-BY-SA 4.0.

Disclaimer

Capability assessments change in weeks, not years. The atlas reflects the state of the art as of June 2026. Where two credible sources disagree, both are noted. Where a claim depends on a still-unreleased model, the verdict is held at the most recent shipped capability.