{
  "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": 24,
  "pageCount": 163,
  "prevPage": 23,
  "nextPage": 25,
  "slideType": "industry_trends",
  "function": "diagnose",
  "density": "overcrowded",
  "nDataPoints": 2,
  "notes": "The slide references the 'Lost in the Middle' research paper by Samaya.ai, UC Berkeley, Stanford, and LMSYS.org.",
  "elementsJson": [
    "headline_text",
    "action_title",
    "paragraph",
    "bullet_list",
    "line_chart"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-06b04d219fea/24",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea#slide-24",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "The researchers found that the models' performance was better when the relevant information for the task occurred in the beginning or in the end of the input, with a more or less dramatic dip in the middle.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a222-49837cdbfbca",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.25,
        "x": 0.41,
        "y": 0.55
      },
      "kind": "chart",
      "text": "Performance by position of document",
      "attrs": null,
      "subkind": "line",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "71e76d8f-f938-4bff-818d-8e43165c4821",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.3,
        "x": 0.67,
        "y": 0.55
      },
      "kind": "chart",
      "text": "Performance by context length",
      "attrs": null,
      "subkind": "line",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "a0a98ebf-4b13-4d17-9ae1-fad9cdd108e4",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.12,
        "w": 0.95,
        "x": 0.02,
        "y": 0.38
      },
      "kind": "list",
      "text": "The researchers found that the models' performance was better when the relevant information for the task occurred in the beginning or in the end of the input, with a more or less dramatic dip in the middle depending on the model. They also found that model performance decreased as input length increased.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "4938d292-81d3-48a7-a2a2-57a7f428de2e",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.38,
        "x": 0.02,
        "y": 0.52
      },
      "kind": "list",
      "text": "The researchers examined the performance of open models MPT-30B-Instruct (8K-token length) and LongChat-13B (16K), and closed ones gpt-3.5 (16K) Claude 1.3 (8K) and Claude 1.3-100K. They found that proprietary models struggled less than open ones.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ace51ae1-72bd-4c5f-913c-5c66da6b6f56",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Accuracy",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a222-4d2b1b6e4617",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.95,
        "x": 0.02,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "The race to the highest context length relies on the hypothesis that a larger context length will result in improved performance for downstream tasks. Research from Samaya.ai, UC Berkeley, Stanford, and LMSYS.org calls this hypothesis into question: When input length is long, even the best available language models can fail on some multi-document question answering and key-value retrieval tasks.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "00d5531d-3211-4d0e-8570-ddc9d531092c",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.8,
        "x": 0.02,
        "y": 0.13
      },
      "kind": "title",
      "text": "Lost in the Middle: long contexts (mostly) don't live up to the expectations",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ec7d428f-f50b-49ce-983c-c3d4780580b2",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a957-3403e5201ed3",
      "evidence": "Title states verdict: 'long contexts (mostly) don't live up to the expectations'.",
      "confidence": 85
    }
  ],
  "frameworks": [
    {
      "name": "Lost in the Middle",
      "slug": null,
      "matchId": "1b5d4d1c-11a9-43a1-a08d-fa37d2b74796",
      "evidence": "Research paper title and findings cited on slide",
      "confidence": 1
    }
  ],
  "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": 31,
      "from": 22,
      "name": "Zoom In",
      "slug": "06-zoom-in",
      "bestFor": "Technical deep-dives, case studies, detailed analysis",
      "matchId": "019dd95a-07fe-70ce-8d3d-ea04ad01cb62",
      "evidence": "Slides drill from emergent-capability mirage into context-length, memory and data-exhaustion details.",
      "position": 3,
      "objective": "Zoom into the data and context-length bottleneck",
      "structure": "The Big Picture -> Key Area of Focus -> Specific Detail -> Implication",
      "confidence": 72,
      "description": "Start broad, then progressively focus on specific details that prove your point"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}