{
  "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": 48,
  "pageCount": 163,
  "prevPage": 47,
  "nextPage": 49,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "dense",
  "nDataPoints": 0,
  "notes": "The slide highlights a specific research finding regarding emergent map-building in blind navigation agents.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "infographic",
    "logo_grid"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-06b04d219fea/48",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea#slide-48",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Map-building is an emergent phenomenon in the course of AI agents learning to navigate.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a224-a4dacc02c846",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.5,
        "w": 0.3,
        "x": 0.65,
        "y": 0.33
      },
      "kind": "image",
      "text": "Visual diagrams (A), (B), and (C) showing navigation paths, agent movement, and collision detection.",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "60bb4285-cf3a-433a-9705-38fa459cd2e1",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.5,
        "w": 0.6,
        "x": 0.03,
        "y": 0.33
      },
      "kind": "list",
      "text": "The Emergence of Maps in the Memories of Blind Navigation Agents shows that giving an agent knowledge of only ego-motion and goal location is sufficient to successfully navigate to the goal. Note that this agent does not have any visual information as input and yet its success rates compared to ‘sighted’ agents are very similar, only efficiency differs.\nThe model doesn’t have any inductive bias towards mapping and is trained with on-policy reinforcement learning. The only mechanism that explains this ability is the memory of the LSTM.\nIt is possible to reconstruct metric maps and detect collisions solely from the hidden state of this agent.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "56de3cb0-49f8-4230-84f0-4dcf8c57dade",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.9,
        "x": 0.03,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "Map-building is an emergent phenomenon in the course of AI agents learning to navigate. It explains why we can feed neural networks images with no explicit maps and can predict navigation policies.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "e9f7e775-00c7-4286-919e-e9cd74f6b360",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.7,
        "x": 0.03,
        "y": 0.14
      },
      "kind": "title",
      "text": "The emergence of maps in the memories of blind navigation agents",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "854d966f-05d6-4042-b086-5ab3610c14bd",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [],
  "frameworks": [],
  "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": 49,
      "from": 41,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-f0e6f3668fb1",
      "evidence": "VLM, autonomous driving, PaLM-E, RT-2, RoboCat, drone-racing, CICERO stack as parallel evidence.",
      "position": 5,
      "objective": "Multimodal models extend from vision-language to embodied robots",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 75,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}