{
  "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": 15,
  "pageCount": 188,
  "prevPage": 14,
  "nextPage": 16,
  "slideType": "industry_trends",
  "function": "present_solution",
  "density": "overcrowded",
  "nDataPoints": 36,
  "notes": "Includes a technical diagram of the Perceiver architecture and a performance comparison table against BERT on the GLUE benchmark.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "flowchart",
    "data_table"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-fd5012384ae3/15",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3#slide-15",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "DeepMind's Perceiver is one such architecture. It solves the Transformers' quadratic dependence on the input length by computing attention between the input and a low-dimensional learnable vector, rather than between the input and itself.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d24-79a44235d772",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.3,
        "w": 0.35,
        "x": 0.58,
        "y": 0.35
      },
      "kind": "diagram",
      "text": "Perceiver architecture diagram showing Input array, Latent array, and Output query array processing.",
      "attrs": null,
      "subkind": "process",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "76135d94-fbde-46c4-a387-ab55e33c6c75",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.6,
        "w": 0.5,
        "x": 0.05,
        "y": 0.2
      },
      "kind": "list",
      "text": "DeepMind's Perceiver is one such architecture. It solves the Transformers' quadratic dependence on the input length by computing attention between the input and a low-dimensional learnable vector, rather than between the input and itself.\nAnother important benefit of Perceiver is its general purpose. It doesn't use domain-specific assumptions and can handle arbitrary input types: images, videos, point clouds, etc.\nPerceiver performs on par with other application-specific architectures, e.g. ViTs for image classification.\nPerceiver IO is an improvement of Perceiver which handles both arbitrary inputs and outputs of any size. This extends Perceiver's capabilities to NLP, games, video generation, etc.\nOn NLP tasks, Perceiver IO doesn't require prior tokenization and directly operates on bytes instead. It still matches the performance of the Transformer-based BERT on GLUE.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "e5d4380f-c086-41bd-badc-be54781141b5",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "GLUE benchmark score: 81.8",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d24-7f1cfbae987b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.03,
        "w": 0.38,
        "x": 0.58,
        "y": 0.9
      },
      "kind": "source-note",
      "text": "Table: Perceiver IO on language: results on the GLUE benchmark (higher is better).",
      "attrs": null,
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "cf41a9b6-9c22-4c77-95de-d7096e9362f1",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.2,
        "w": 0.38,
        "x": 0.58,
        "y": 0.7
      },
      "kind": "table",
      "text": "Model performance comparison table on GLUE benchmark.",
      "attrs": null,
      "subkind": "data",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "8cc4bc64-7dc7-43bc-a905-9a6409150649",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.7,
        "x": 0.05,
        "y": 0.13
      },
      "kind": "title",
      "text": "Transformers extend into efficient self-attention-based architectures",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "79296364-2d6d-4607-a8df-d56807fa0ad8",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [],
  "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": 18,
      "from": 11,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-45052222b756",
      "evidence": "Multiple evidence slides (vision, audio, 3D, MLPs comeback) all converging on transformer dominance.",
      "position": 2,
      "objective": "Show transformers spreading across AI domains",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 82,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}