{
  "docId": "019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "docSlug": "bf350dd574c19997",
  "documentTitle": "2021 Air Street Capital The State of AI Report 2021",
  "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": 36,
  "pageCount": 188,
  "prevPage": 35,
  "nextPage": 37,
  "slideType": "industry_trends",
  "function": "summarize",
  "density": "dense",
  "nDataPoints": 1,
  "notes": "The slide explains the technical shift from GANs to diffusion models in AI research.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "photo"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-fd5012384ae3/36",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3#slide-36",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Diffusion models' training is more stable than GAN's and outperforms them on several well-established datasets in image generation, audio synthesis, shape generation and music generation.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d27-30f4388309bf",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.5,
        "w": 0.4,
        "x": 0.58,
        "y": 0.35
      },
      "kind": "image",
      "text": "Visual representation of noise addition process and generated samples",
      "attrs": null,
      "subkind": "illustration",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "97683de0-cb97-44f4-a853-e81fd231b11b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.55,
        "w": 0.55,
        "x": 0.02,
        "y": 0.33
      },
      "kind": "list",
      "text": "Principle: Given an image from a dataset D, after enough steps of random noise additions, we approximately end up with a sample from the distribution of the noise. What if it was possible to revert the process and recover an image from the distribution of the dataset D by sampling from noise?\nMethod: Diffusion models solve this problem by modeling the inverse distribution (generating denoised images from noisy ones) at each step as a Gaussian whose mean and covariance are parametrized as a DNN.\nDiffusion models are not new, but recent improvements have made them theoretically and practically appealing.\nAlthough they are slower, they beat GANs on ImageNet across all resolutions from 64x64 to 512x512.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "b6a39236-88ab-45c6-9e78-a87e069cd6a7",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.1,
        "w": 0.55,
        "x": 0.02,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "Diffusion models' training is more stable than GAN's and outperforms them on several well-established datasets in image generation, audio synthesis, shape generation and music generation.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "eb0c4e49-ccfc-465a-b5f9-27a4e4b4a7be",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.03,
        "w": 0.4,
        "x": 0.58,
        "y": 0.85
      },
      "kind": "source-note",
      "text": "Figure 1: Selected samples from our best ImageNet 512x512 model (FID 3.85)",
      "attrs": null,
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "9bcf3e45-4c86-4122-8929-8547095472a6",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.55,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "GANs have a serious new adversary: diffusion models",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f4edf58c-f240-4f26-ba05-4ce0c268d2b9",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Contrast Pairs",
      "slug": "contrast-pairs",
      "agent": "Storyteller",
      "layer": "loop",
      "matchId": "019dd95a-0fd5-7148-8ed0-f4b71e411f2c",
      "evidence": "GANs vs diffusion contrast structure.",
      "confidence": 75
    },
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-0fd5-7148-8ed0-f2aae36c1103",
      "evidence": "Title 'GANs have a serious new adversary: diffusion models' = clear finding.",
      "confidence": 85
    },
    {
      "name": "Comparison frame",
      "slug": "comparison-frame",
      "agent": null,
      "layer": "slide",
      "matchId": "ff91dc88-7f20-40dd-b12b-b35d506ad9b4",
      "evidence": "The slide compares the performance of diffusion models with GANs, highlighting their advantages.",
      "confidence": 0.7
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 152,
      "from": 5,
      "beatId": "019dd95a-0682-776c-8e35-0affbadec38e",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Definitions p.5-6, Exec Summary p.7, then Sections 1-3 catalog evidence.",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 152,
      "from": 10,
      "beatId": "019dd95a-0682-776c-8e35-1b3ee3a1c012",
      "arcName": "The Mountain",
      "arcSlug": "mountain",
      "beatName": "Rising Action",
      "beatSlug": "mountain-rising-action",
      "evidence": "Three sections accumulating evidence of accelerating progress.",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 40,
      "from": 36,
      "name": "Aha Moment",
      "slug": "03-aha-moment",
      "bestFor": "Data-heavy sections, research findings, analytical arguments",
      "matchId": "019dd95a-07fe-70ce-8d3d-53192fc75199",
      "evidence": "Diffusion models > GANs, then CLIP, then DALL-E payoff.",
      "position": 5,
      "objective": "Reveal multimodal generative AI breakthrough",
      "structure": "The Problem/Question -> What the Data Shows -> The Insight",
      "confidence": 72,
      "description": "Establish a tension, present data, then deliver the insight that changes everything"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}