{
  "docId": "019dd923-5e88-73ef-bd5c-f6fdd8f57895",
  "docSlug": "3a892393f72c1ff5",
  "documentTitle": "2019 Air Street Capital The State of AI Report 2019",
  "authorId": "AirStreetCapital",
  "authorName": "Nathan Benaich and Ian Hogarth",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.777,
  "pageNumber": 11,
  "pageCount": 136,
  "prevPage": 10,
  "nextPage": 12,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "overcrowded",
  "nDataPoints": 1,
  "notes": "The slide highlights the transition from supervised learning to multi-agent reinforcement learning.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "scatter_plot",
    "paragraph"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-f6fdd8f57895/11",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-f6fdd8f57895",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-f6fdd8f57895.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-f6fdd8f57895#slide-11",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "The market cost of the compute resource to train AlphaStar has been estimated at $26M.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-41ab-72d9-bbc3-8ba8cb096397",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.35,
        "x": 0.62,
        "y": 0.4
      },
      "kind": "chart",
      "text": "AlphaStar Training League progress chart showing MMR increase over training days.",
      "attrs": null,
      "subkind": "scatter",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "58b53082-a435-4c5e-a4f2-be0e5e962e70",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.6,
        "x": 0.02,
        "y": 0.45
      },
      "kind": "list",
      "text": "AlphaStar was first trained by supervised learning on a set of human games. After this, AlphaStar's novel approach used a multi-agent training algorithm that effectively created a league of agents competing against each other and collectively exploring the huge strategic space. The final AlphaStar agent is produced by Nash averaging, which combines the most effective mix of strategies developed by individual agents.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "116c0904-3a39-489b-b53a-943d428ba356",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.12,
        "w": 0.6,
        "x": 0.02,
        "y": 0.32
      },
      "kind": "list",
      "text": "DeepMind beat a world class player 5-0. StarCraft II still cannot be considered to be 'solved' due to various constraints on the action space. This is nonetheless a major breakthrough.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "78c4898f-ae2c-41b7-919f-3a3e1bc34d09",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.6,
        "x": 0.02,
        "y": 0.72
      },
      "kind": "list",
      "text": "The market cost of the compute resource to train AlphaStar has been estimated at $26M. The cost of the world class team at DeepMind's working on AlphaStar could be similar.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f35bd967-30b4-490e-bc65-315f693ee316",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Estimated MMR: $26M",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-41ab-72d9-bbc3-8e1c76a5c873",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.6,
        "x": 0.02,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "StarCraft integrates various hard challenges for ML systems: Operating with imperfect information, controlling a large action space in real time and making strategic decisions over a long time horizon.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "9cdd98f8-af90-4c6d-bbf4-a8c872663b7b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.4,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "RL conquers new territory: StarCraft II",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "b6330f40-56f1-4707-9e1c-a432a56409c6",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-0fd5-7148-8ecc-a940f5e3d1ad",
      "evidence": "Title quantifies AlphaStar compute cost at $26M.",
      "confidence": 80
    },
    {
      "name": "Credibility Transfer",
      "slug": "credibility-transfer",
      "agent": "Storyteller",
      "layer": "slide",
      "matchId": "019dd95a-0fd5-7148-8ecc-aee411426aa3",
      "evidence": "DeepMind StarCraft cited as authority example.",
      "confidence": 65
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 92,
      "from": 5,
      "beatId": "019dd95a-0682-776c-8e34-ce57e5bbe574",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Definitions through Research/Talent/Industry — fact-dense case studies and benchmarks.",
      "position": 1,
      "confidence": 70,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 92,
      "from": 9,
      "beatId": "019dd95a-0682-776c-8e34-ddbccb391af1",
      "arcName": "Voyage and Return",
      "arcSlug": "voyage-return",
      "beatName": "The Unknown",
      "beatSlug": "voyage-return-the-unknown",
      "evidence": "Research/Talent/Industry frontiers explored.",
      "position": 2,
      "confidence": 45,
      "parentBeatName": "Complication",
      "parentBeatSlug": "complication"
    }
  ],
  "loops": [
    {
      "to": 15,
      "from": 10,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3c-9ac033d2887b",
      "evidence": "Six case studies (Montezuma, StarCraft II, CTF, Dota 99.4%, robot dexterity) all titled 'RL conquers new territory'.",
      "position": 2,
      "objective": "Show RL is conquering new domains via stacked breakthrough examples",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 80,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}