{
  "docId": "019dd923-5e88-73ef-bd5d-06b04d219fea",
  "docSlug": "dd91c78f6570bf29",
  "documentTitle": "2023 Air Street Capital The State of AI Report 2023",
  "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": 55,
  "pageCount": 163,
  "prevPage": 54,
  "nextPage": 56,
  "slideType": "industry_trends",
  "function": "summarize",
  "density": "overcrowded",
  "nDataPoints": 4,
  "notes": "Includes a diagram of the model architecture showing prompt inputs and image inputs leading to a valid mask output.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "process_diagram",
    "photo"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-06b04d219fea/55",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea#slide-55",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Meta tested SAM on 23 out of domain image datasets outperforming existing SoTA on 70%+ of cases.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a225-34b973986694",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.25,
        "x": 0.7,
        "y": 0.53
      },
      "kind": "diagram",
      "text": "Model architecture diagram showing segmentation prompt and image inputs to a model producing a valid mask.",
      "attrs": null,
      "subkind": "process",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "7a50a287-e174-4d1b-8176-7c446935af61",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.95,
        "x": 0.02,
        "y": 0.38
      },
      "kind": "list",
      "text": "Taking inspiration from large language models which are pre-trained on vast datasets and exhibit zero-shot capabilities via prompting, Meta researchers set out to build a model that enables general promptable segmentation: given any prompt, the model should be able to identify and segment any object in any image.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "0b68de8d-97b0-4713-a202-4880c9c0569f",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.95,
        "x": 0.02,
        "y": 0.22
      },
      "kind": "list",
      "text": "Meta introduced a large-scale project called “Segment Anything” which included the release of 1B segmentation masks on a 11M image dataset (SA-1B), and a segmentation model (SAM) with an Apache 2.0 commercial use license. Meta tested SAM on 23 out of domain image datasets outperforming existing SoTA on 70%+ of cases.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ca8019e6-b9af-4fd6-9c75-3af19478df8e",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.12,
        "w": 0.65,
        "x": 0.02,
        "y": 0.79
      },
      "kind": "list",
      "text": "Through prompt engineering, SAM can be applied to other tasks including edge detection, object proposal generation, and instance segmentation and preliminary results were shown combining SAM + CLIP for text prompts.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f55efdfb-9d3d-4873-a737-2fcea7f01183",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.65,
        "x": 0.02,
        "y": 0.54
      },
      "kind": "list",
      "text": "The model has two components: (i) An heavyweight encoder (ViT) to compute a one-time image embedding, (ii) a lightweight interactive module (that can run on CPU in a browser) consisting of a prompt encoder that embeds the user prompt, and mask decoder that predicts the segmentation masks.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ff5c8d0a-e610-4448-9b36-41c768c69933",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.65,
        "x": 0.02,
        "y": 0.7
      },
      "kind": "list",
      "text": "A model-in-the-loop data-engine was used to generate the training data, with the final SA-1B generated entirely automatically by applying SAM.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ff7b87ac-9abc-41e9-a874-1488850e7299",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "segmentation masks: 1B",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a225-38df7e72e9c2",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.9,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "Segment Anything: a promptable segmentation model with zero-shot generalisation",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "e425259c-3e19-40b5-9e69-1390f5ac9c16",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Concrete Language",
      "slug": "concrete-language",
      "agent": "Storyteller",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a957-5aa3c4b98d61",
      "evidence": "'23 out of domain image datasets, SoTA on 70%+ of cases' specifics.",
      "confidence": 70
    },
    {
      "name": "NxN matrix",
      "slug": "matrix-nxn",
      "agent": null,
      "layer": "slide",
      "matchId": "8807de22-d58b-4f92-aede-85e5f0198c33",
      "evidence": "diagram/process: Model architecture diagram showing segmentation prompt and image inputs to a model producing a valid mask.",
      "confidence": 0.6
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 120,
      "from": 11,
      "beatId": "019dd95a-0682-776c-8e35-41afd44ef59f",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Research + Industry sections inventory model, compute, funding facts.",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 120,
      "from": 11,
      "beatId": "019dd95a-0682-776c-8e35-523bfb7f96e6",
      "arcName": "The Mountain",
      "arcSlug": "mountain",
      "beatName": "Rising Action",
      "beatSlug": "mountain-rising-action",
      "evidence": "Escalating capabilities, compute concentration and capital flows.",
      "position": 2,
      "confidence": 45,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 58,
      "from": 50,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-f70f6435b722",
      "evidence": "Text-to-video, instruction editing, Gaussian splatting, NeRFs, depth, SAM, DINOv2, weather, music.",
      "position": 6,
      "objective": "Generative media and visual foundation models proliferate",
      "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
}