{
  "docId": "019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "docSlug": "8757f1b44ef7f176",
  "documentTitle": "2024 Air Street Capital The State of AI Report 2024",
  "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": 198,
  "pageCount": 213,
  "prevPage": 197,
  "nextPage": 199,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "dense",
  "nDataPoints": 0,
  "notes": "The slide showcases a specific technical breakthrough in AI interpretability, using the 'Golden Gate Bridge' feature as a primary example.",
  "elementsJson": [
    "paragraph",
    "bubble_chart",
    "screenshot",
    "photo"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0856e1444fb9/198",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9#slide-198",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Anthropic's interpretability team used sparse autoencoders - neural networks that learn efficient representations of data by emphasizing important features and ensuring only a few are active at any one time - to decompose activations of Claude 3 Sonnet into interpretable components.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d23-eb37b55fdcb9",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.42,
        "x": 0.06,
        "y": 0.43
      },
      "kind": "image",
      "text": "Visualization of feature clusters related to inner conflict and guilt.",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "22233bb4-9861-4165-9412-5ccd8639669a",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.6,
        "x": 0.2,
        "y": 0.8
      },
      "kind": "image",
      "text": "Example of feature activation related to biological weapons.",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ac7c4463-78e3-4f50-b859-80824abcc631",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.44,
        "x": 0.52,
        "y": 0.43
      },
      "kind": "image",
      "text": "Golden Gate Bridge feature activation examples in text and images.",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "fde7d5cc-0141-4f1e-a957-2ed3915bc785",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.96,
        "x": 0.02,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "Anthropic's interpretability team used sparse autoencoders - neural networks that learn efficient representations of data by emphasizing important features and ensuring only a few are active at any one time - to decompose activations of Claude 3 Sonnet into interpretable components. They also showed that by 'pinning' a feature to 'active' you could control the output - famously turning up the strength of the Golden Gate feature.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "e2c4d1ce-a5f5-4b5a-beab-028eafd90ae9",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.04,
        "w": 0.4,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "Anthropic breaks open the black box...",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "7e080e09-ac15-4d34-a7bc-58baf9e28ea4",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a95a-c86ff5eff123",
      "evidence": "Title 'Anthropic breaks open the black box...'",
      "confidence": 80
    },
    {
      "name": "Metaphor & Analogy",
      "slug": "metaphor-analogy",
      "agent": "Storyteller",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a95a-ccf8344e9384",
      "evidence": "'Black box' metaphor for model interpretability",
      "confidence": 70
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 204,
      "from": 154,
      "beatId": "019dd95a-0682-776c-8e35-605159a069e6",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Implications (So What)",
      "beatSlug": "triple-take-the-implications-so-what",
      "evidence": "Politics + Safety sections explore consequences and risks",
      "position": 2,
      "confidence": 78,
      "parentBeatName": "Reflection",
      "parentBeatSlug": "reflection"
    },
    {
      "to": 204,
      "from": 175,
      "beatId": "019dd95a-0682-776c-8e35-74f05e198faf",
      "arcName": "Voyage and Return",
      "arcSlug": "voyage-return",
      "beatName": "The Return",
      "beatSlug": "voyage-return-the-return",
      "evidence": "Safety section returns to risks/governance debate",
      "position": 4,
      "confidence": 55,
      "parentBeatName": "Resolution",
      "parentBeatSlug": "resolution"
    }
  ],
  "loops": [
    {
      "to": 201,
      "from": 187,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3e-79a3afd2ac42",
      "evidence": "Emergent capabilities, sycophancy, DPO, hybrid RLHF, critiques, transparency, reward tampering, SAEs",
      "position": 23,
      "objective": "Catalogue alignment, RLHF, sycophancy and interpretability research",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 76,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}