{
  "docId": "019dd923-5e88-73ef-bd5c-f812573a947a",
  "docSlug": "eea7524c557036f4",
  "documentTitle": "2020 Air Street Capital The State of AI Report 2020",
  "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": 28,
  "pageCount": 177,
  "prevPage": 27,
  "nextPage": 29,
  "slideType": "industry_trends",
  "function": "analyze_data",
  "density": "dense",
  "nDataPoints": 12,
  "notes": "The slide uses a grouped bar chart to compare performance across three categories (Commonsense, Linguistics, Knowledge) for four model sizes (Small, Medium, Large, X-Large).",
  "elementsJson": [
    "headline_text",
    "paragraph",
    "bullet_list",
    "bar_chart_grouped",
    "footnote"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-f812573a947a/28",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-f812573a947a",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-f812573a947a.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-f812573a947a#slide-28",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "GPT-3's performance is lopsided with large knowledge gaps.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c5-73ac-aa61-c253b57d862c",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.55,
        "w": 0.38,
        "x": 0.58,
        "y": 0.32
      },
      "kind": "chart",
      "text": "GPT-3 Few Shot Test Performance",
      "attrs": null,
      "subkind": "bar-grouped",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "22bbd888-d51d-49cd-afec-52dbb7f586e9",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.55,
        "w": 0.52,
        "x": 0.02,
        "y": 0.32
      },
      "kind": "list",
      "text": "Large models like GPT-3 that are pre-trained on vast language corpora obviate the need for task-specific fine-tuning on a specific dataset. This enables few-shot learning on new tasks. A new benchmark measures knowledge acquired during pre-training by evaluating in few-shot settings (% avg. weighted accuracy below).\nWhile the GPT-3 X-Large improves over random chance by over 20 percentage points on average, the model's accuracy ranges from 69% for US Foreign Policy to 26% for College Chemistry. Moreover, GPT-3's average confidence is a poor estimator of its accuracy and can be off by up to 24%.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "b3949038-da74-4f78-ba4c-411c79538189",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Weighted accuracy: 69%",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c5-73ac-aa61-c6ad4c9ab7a0",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.95,
        "x": 0.02,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "A multi-task language understanding challenge tests for world knowledge and problem solving ability across 57 tasks including maths, US history, law and more. GPT-3's performance is lopsided with large knowledge gaps.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "5c773ab1-e9ca-489f-9a06-f2236f657958",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.04,
        "w": 0.6,
        "x": 0.13,
        "y": 0.93
      },
      "kind": "source-note",
      "text": "Figure note: “Small” (2.7B parameters), “Medium” (6.7B), “Large” (13B) and “X-Large” (175B).",
      "attrs": null,
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "fea84cf5-1bb8-4d84-b896-bd2e23c41e30",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.95,
        "x": 0.02,
        "y": 0.13
      },
      "kind": "title",
      "text": "What’s next after SuperGLUE? More challenging NLP benchmarks zero-in on knowledge",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "40591bf6-4be3-4d64-adf6-f920b3cbea2e",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "So What? Test",
      "slug": "so-what-test",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-0fd5-7148-8ece-917e74bb800a",
      "evidence": "Callout: GPT-3 has knowledge gaps.",
      "confidence": 70
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 129,
      "from": 10,
      "beatId": "019dd95a-0682-776c-8e34-ed2a8f38a754",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Research, Talent, Industry sections inventory what happened in AI in 2020.",
      "position": 1,
      "confidence": 60,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 129,
      "from": 10,
      "beatId": "019dd95a-0682-776c-8e34-fc384f439bba",
      "arcName": "The Mountain",
      "arcSlug": "mountain",
      "beatName": "Rising Action",
      "beatSlug": "mountain-rising-action",
      "evidence": "Research, Talent and Industry sections build momentum of AI progress.",
      "position": 2,
      "confidence": 40,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 29,
      "from": 24,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3c-f35d665a8850",
      "evidence": "Sequence of GPT-3/T5/code translation/DrRepair examples.",
      "position": 3,
      "objective": "Show transformers unlocking new NLP/code/vision use cases",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 70,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}