{
  "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": 66,
  "pageCount": 213,
  "prevPage": 65,
  "nextPage": 67,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "balanced",
  "nDataPoints": 0,
  "notes": "The slide contrasts Intelligent Go-Explore (using LLMs) with Classic Go-Explore (using hand-crafted heuristics) through a process diagram.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "process_diagram"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0856e1444fb9/66",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9#slide-66",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "Switching from GPT-4 to GPT-3.5 resulted in a significant performance drop across all environments, suggesting that IGE's performance scales with the capabilities of the underlying language model.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d14-476b52264277",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.38,
        "x": 0.6,
        "y": 0.45
      },
      "kind": "diagram",
      "text": "Comparison diagram between Intelligent Go-Explore and Classic Go-Explore showing state selection, action choice, and archive updating processes.",
      "attrs": null,
      "subkind": "process",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "1c5fd9a1-3e5b-4652-a0c9-301477f1e192",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.65,
        "w": 0.55,
        "x": 0.03,
        "y": 0.22
      },
      "kind": "list",
      "text": "In 2019, Uber published Go-Explore, an RL agent that solved hard-exploration problems by archiving discovered states and iteratively returning to and exploring from promising ones. In 2024, LLMs are supercharging it.\nIntelligent Go-Explore (IGE) uses an LLM to guide state selection, action choice, and archive updating, rather than the original Go-Explore's hand-crafted heuristics. This enabled more flexible and intelligent exploration in complex environments.\nThis approach also allowed IGE to recognize and capitalize on promising discoveries, a key aspect of open-ended learning systems.\nIt significantly outperformed other LLM agents on mathematical reasoning, grid worlds, and text-based adventure games.\nSwitching from GPT-4 to GPT-3.5 resulted in a significant performance drop across all environments, suggesting that IGE's performance scales with the capabilities of the underlying language model.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ef813032-f6c9-4c31-b7e4-1a8d824a6ef8",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.4,
        "x": 0.03,
        "y": 0.14
      },
      "kind": "title",
      "text": "...while LLMs improve RL performance",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "eafe21fc-33b5-455c-81a5-3eb60ab65ca6",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Chiasmus",
      "slug": "chiasmus",
      "agent": "Storyteller",
      "layer": "loop",
      "matchId": "019dd95a-1055-74e0-a959-2667989aad7a",
      "evidence": "Two consecutive titles flip A→B and B→A structure",
      "confidence": 80
    },
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a959-21924b027b5c",
      "evidence": "Title '...while LLMs improve RL performance' completes the pair",
      "confidence": 85
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 153,
      "from": 9,
      "beatId": "019dd95a-0682-776c-8e35-5f2398b8d1d0",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Exec summary + Research + Industry sections inventory the year",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 85,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-6d9d8ee06d0a",
      "arcName": "Voyage and Return",
      "arcSlug": "voyage-return",
      "beatName": "The Unknown",
      "beatSlug": "voyage-return-the-unknown",
      "evidence": "Research section explores frontier model uncharted territory",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Complication",
      "parentBeatSlug": "complication"
    }
  ],
  "loops": [
    {
      "to": 71,
      "from": 59,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3e-3f24b9b4c1e3",
      "evidence": "ARC challenge, planning failures, COT, open-endedness, MCTS, FunSearch, RL-LLM combos",
      "position": 8,
      "objective": "Catalogue progress and limits in reasoning 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
}