{
  "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": 41,
  "pageCount": 213,
  "prevPage": 40,
  "nextPage": 42,
  "slideType": "industry_trends",
  "function": "summarize",
  "density": "dense",
  "nDataPoints": 0,
  "notes": "The slide explains technical improvements in AI image generation models (Stability AI).",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "paragraph",
    "process_diagram"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0856e1444fb9/41",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9#slide-41",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Adversarial diffusion distillation speeds up image generation by reducing the sampling steps needed to create high-quality images from potentially hundreds down to 1-4, while maintaining high fidelity.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d11-337a76b01914",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.28,
        "x": 0.7,
        "y": 0.45
      },
      "kind": "image",
      "text": "Technical diagram showing the adversarial diffusion distillation process with student and teacher models.",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "005dc02f-9430-44a7-b9c9-45c71b0bb4d5",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.6,
        "w": 0.65,
        "x": 0.03,
        "y": 0.31
      },
      "kind": "list",
      "text": "Adversarial diffusion distillation speeds up image generation by reducing the sampling steps needed to create high-quality images from potentially hundreds down to 1-4, while maintaining high fidelity.\nIt combines adversarial training with score distillation: the model is trained just using a pre-trained diffusion model as a guide.\nAs well as unlocking single-step generation, the authors focused on reducing computational complexity and improving sampling efficiency.\nRectified flow improves upon traditional diffusion methods by connecting data and noise through a direct, straight line, rather than a curved path.\nThey combined this with a novel transformer-based architecture for text-to-image that allow for a bidirectional flow of information between text and image components. This enhances the model's ability to generate more accurate and coherent high-resolution images based on textual descriptions.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "be8af126-a401-48ae-b6e3-78e1703a8020",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.95,
        "x": 0.03,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "Moving on from diffusion models for text-to-image, Stability AI have continued to search for refinements that increase quality while bringing about greater efficiency.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "571d97a4-661f-463e-a03a-b90f38490512",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.8,
        "x": 0.03,
        "y": 0.13
      },
      "kind": "title",
      "text": "Diffusion models for image generation become more and more sophisticated",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "55161835-1bd6-49b0-9c05-d9fc765714a3",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [],
  "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": 44,
      "from": 40,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3e-3752140f3c2d",
      "evidence": "VLMs, diffusion image, Stable Video Diffusion, gated text-to-video, Movie Gen audio",
      "position": 6,
      "objective": "Map progress in multimodal vision/video/audio generation",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 76,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}