{
  "docId": "019dd923-5e88-73ef-bd5d-0d0f98caffe1",
  "docSlug": "5df6bd1b0447b5f6",
  "documentTitle": "2025 Air Street Capital The State of AI Report 2025",
  "authorId": "AirStreetCapital",
  "authorName": "Air Street Capital",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.777,
  "pageNumber": 75,
  "pageCount": 313,
  "prevPage": 74,
  "nextPage": 76,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "dense",
  "nDataPoints": 2,
  "notes": "The slide summarizes research findings on brain-model convergence, highlighting that early layers align with sensory cortices while later layers require more training to align with prefrontal regions.",
  "elementsJson": [
    "headline_text",
    "paragraph",
    "bullet_list",
    "line_chart"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0d0f98caffe1/75",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1#slide-75",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Brain-like representations emerge progressively during training.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-41ac-7180-a820-27125c4c27af",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.2,
        "w": 0.25,
        "x": 0.72,
        "y": 0.7
      },
      "kind": "chart",
      "text": "MEG alignment over training steps",
      "attrs": null,
      "subkind": "line",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "3740b71b-3f37-4f2d-9775-a08e7b5ee826",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.2,
        "w": 0.25,
        "x": 0.72,
        "y": 0.48
      },
      "kind": "chart",
      "text": "fMRI alignment over training steps",
      "attrs": null,
      "subkind": "line",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f7c73e58-a174-4e46-b639-6e3260ea2c30",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.65,
        "x": 0.02,
        "y": 0.47
      },
      "kind": "list",
      "text": "fMRI (8 subjects, ~10,000 images each, for high-res spatial maps of cortical activity) and MEG (4 subjects, ~22,500 images each, for high-res temporal dynamics) recordings were compared against DINOv3 activations.\nThree metrics of brain-model similarity were assessed: encoding score (linear similarity), spatial score (layer ↔ cortical hierarchy), and temporal score (layer ↔ brain response timing).\nBrain-like representations emerge progressively during training. Early visual regions and fast MEG responses align quickly, while prefrontal cortices and late temporal windows require far more training, closely echoing the developmental trajectory of the human cortex.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "d0856e7a-bc68-405e-9c82-ac574f19003b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Brain-model similarity score",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-41ac-7180-a820-2bb4a4a89d82",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.65,
        "x": 0.02,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "By systematically varying model size, training scale, and image type in DINOv3 (Meta's latest self-supervised image model trained on billions of images), researchers show that brain-model convergence emerges in a specific sequence. They find that early layers align with sensory cortices, while only prolonged training and human-centric data drive alignment with prefrontal regions. Larger models converge faster, and later-emerging representations mirror cortical properties like expansion, thickness, and slow timescales.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "6c44479c-8038-47d5-a82c-b69dbaba9d2f",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.7,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "Can vision models align with human brains...and how does that alignment emerge?",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "a9304baf-d7e6-4b29-a727-48bafeb960b1",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Chart Selection Guide",
      "slug": "chart-selection-guide",
      "agent": "Designer",
      "layer": "slide",
      "matchId": "25b4eb09-a02b-4126-8c09-cf6fc0e6d4eb",
      "evidence": "chart/line: fMRI alignment over training steps",
      "confidence": 0.5
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 189,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-9f380673831f",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Sections 1-2 lay out research findings and industry data with charts and case studies.",
      "position": 1,
      "confidence": 70,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 89,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-ac68557d958b",
      "arcName": "The Onion",
      "arcSlug": "onion",
      "beatName": "First Layer",
      "beatSlug": "onion-first-layer",
      "evidence": "Research section unpacks technical layer beneath the headlines.",
      "position": 2,
      "confidence": 45,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}