{
  "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": 59,
  "pageCount": 188,
  "prevPage": 58,
  "nextPage": 60,
  "slideType": "industry_trends",
  "function": "summarize",
  "density": "dense",
  "nDataPoints": 4,
  "notes": "Includes specific examples of legal NLP (legalBERT/CaseHOLD) and malware detection (SoReL-20M).",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "infographic",
    "logo_grid"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-fd5012384ae3/59",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3#slide-59",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "As data-hungry deep learning conquers more applications, better domain-specific datasets are needed. Legal NLP and malware exemplify this struggle as new pretraining datasets and benchmarks come to the rescue.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d29-f983d6aee9c5",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.38,
        "x": 0.58,
        "y": 0.35
      },
      "kind": "image",
      "text": "Prompt (citing text) / Correct answer (holding statement) / Incorrect answer (holding statement from different context)",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "c7980835-b9f2-43fe-8e94-237e66074bae",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.2,
        "x": 0.75,
        "y": 0.65
      },
      "kind": "image",
      "text": "SOREL 20M",
      "attrs": null,
      "subkind": "logo",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "edb48bb5-b7b5-47d2-ad83-8132687cf96b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.6,
        "w": 0.55,
        "x": 0.03,
        "y": 0.31
      },
      "kind": "list",
      "text": "Legal NLP: Several works have shown that pretraining on existing legal datasets didn't help more than pretraining on general texts when NLP is applied to legal texts. Stanford's RegLab introduces a huge dataset of ~3.5M legal decisions (37GB of text) across American federal courts, on which they pretrain their language model, legalBERT. legalBERT significantly outperforms a general purpose BERT on 3 tasks, including CaseHOLD, a new task consisting of 53,000 Q&As from American Case Law. Malware: SophosAI and ReversingLabs introduced SoReL-20M, the largest dataset for malware detection. It contains 20 million files with significantly more metadata than older datasets. They find that 20 million files is a large enough size to differentiate between machine learning models of different capacities. They also released models trained on this dataset that can serve as baselines.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "1c324045-1b28-4ec8-91dd-8ee9ecefa3d8",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "dataset size: 20M",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d29-fea14d455746",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.95,
        "x": 0.03,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "As data-hungry deep learning conquers more applications, better domain-specific datasets are needed. Legal NLP and malware exemplify this struggle as new pretraining datasets and benchmarks come to the rescue.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "fff6ed25-560a-4d3b-b2d4-ec29467d6f19",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.6,
        "x": 0.03,
        "y": 0.14
      },
      "kind": "title",
      "text": "Better datasets for machine learning in production: legal documents and malware",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "e5283d34-c9ac-49ce-8559-c2b214dd5ce6",
      "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": 63,
      "from": 56,
      "name": "Iceberg",
      "slug": "10-iceberg",
      "bestFor": "Consulting, complex problem solving, organizational change",
      "matchId": "019dd95a-07fe-70ce-8d3d-5f004461126b",
      "evidence": "Code availability, data cascades, dataset documentation, collusion rings, code+conferences.",
      "position": 8,
      "objective": "Diagnose openness and trust crisis in ML research",
      "structure": "The Symptom (Visible) -> The System (Hidden) -> The Root Cause",
      "confidence": 75,
      "description": "Reveal that the visible problem is merely a symptom of a deeper root cause"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}