{
  "docId": "019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "docSlug": "8757f1b44ef7f176",
  "documentTitle": "2024 Air Street Capital The State of AI Report 2024",
  "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": 31,
  "pageCount": 213,
  "prevPage": 30,
  "nextPage": 32,
  "slideType": "industry_trends",
  "function": "diagnose",
  "density": "dense",
  "nDataPoints": 0,
  "notes": "Includes a diagram illustrating the feedback loop of model collapse.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "paragraph",
    "process_diagram",
    "logo_grid"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0856e1444fb9/31",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9#slide-31",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "This creates a “first mover advantage”, as sustained access to diverse, human-generated data will become increasingly critical for maintaining model quality.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d10-32dbca33305d",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.3,
        "w": 0.3,
        "x": 0.65,
        "y": 0.5
      },
      "kind": "diagram",
      "text": "Probable events are overestimated. Improbable events are underestimated. Finite sampling. Approximate fitting. Data^n. Model n. Probable events poison reality. Tails shrink over time.",
      "attrs": null,
      "subkind": "process",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "bc8a7134-8372-4925-aed9-a2b4741da9bb",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.5,
        "w": 0.6,
        "x": 0.03,
        "y": 0.3
      },
      "kind": "list",
      "text": "A Nature paper from Oxford and Cambridge researchers found model collapse occurs across various AI architectures, including fine-tuned language models, challenging the idea that pre-training or periodic exposure to small amounts of original data can prevent degradation (measured by Perplexity score).\nThis creates a “first mover advantage”, as sustained access to diverse, human-generated data will become increasingly critical for maintaining model quality.\nHowever, these results are primarily focused on a scenario where real data is replaced with synthetic data over generations. In practise, real and synthetic data usually accumulates.\nOther research suggests that, provided the proportion of synthetic data doesn’t get too high, collapse can usually be avoided.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "7f7bee7b-7abe-4dd8-8597-29f9fe605e09",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.9,
        "x": 0.03,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "As model builders motor ahead, researchers have focused on trying to assess if there’s a tipping point in the quantity of synthetic data that triggers these kinds of outcomes and if any mitigations work",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "94ec9ac1-f2d0-459a-924a-c27a3fd05995",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.6,
        "x": 0.03,
        "y": 0.14
      },
      "kind": "title",
      "text": "...but Team Model Collapse isn’t going down without a fight",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "7f3c0753-dde1-4280-a5ef-5a2503e8150a",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Callback",
      "slug": "callback",
      "agent": "Storyteller",
      "layer": "loop",
      "matchId": "019dd95a-1055-74e0-a958-acbdec4d8617",
      "evidence": "'...but' explicitly references prior slide",
      "confidence": 70
    },
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a958-ab0f79fc65b2",
      "evidence": "Title '...but Team Model Collapse isn't going down without a fight' uses callback",
      "confidence": 80
    },
    {
      "name": "Cognitive Load",
      "slug": "cognitive-load",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "b7e13b37-f714-4dd0-9304-2faa95dc6e6c",
      "evidence": "Probable events are overestimated. Improbable events are underestimated. Finite sampling. Approximate fitting. Data^n. Model n. Probable events poison reality. Tails shrink over time.",
      "confidence": 0.5
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 153,
      "from": 9,
      "beatId": "019dd95a-0682-776c-8e35-5f2398b8d1d0",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Exec summary + Research + Industry sections inventory the year",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 85,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-6d9d8ee06d0a",
      "arcName": "Voyage and Return",
      "arcSlug": "voyage-return",
      "beatName": "The Unknown",
      "beatSlug": "voyage-return-the-unknown",
      "evidence": "Research section explores frontier model uncharted territory",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Complication",
      "parentBeatSlug": "complication"
    }
  ],
  "loops": [
    {
      "to": 37,
      "from": 30,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3e-2e9bfb3e3bea",
      "evidence": "Synthetic data, FineWeb, embeddings, RAG, power grid, JEST data curation",
      "position": 4,
      "objective": "Survey data, retrieval and training-infra advances",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 72,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}