{
  "docId": "019dd923-5fec-763b-95d0-451a96d1edee",
  "docSlug": "bi-7495499d6a6908b8",
  "documentTitle": "AgileRL raised $7.5M with this pitch deck to help firms train AI",
  "authorId": "Pitchdecks",
  "authorName": "AgileRL",
  "documentKindSlug": "pitchdeck",
  "documentKindLabel": "Pitch deck",
  "sourceTypeSlug": "strategy_consulting",
  "sourceTypeLabel": "Strategy consulting",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.6,
  "pageNumber": 2,
  "pageCount": 12,
  "prevPage": 1,
  "nextPage": 3,
  "slideType": "why_now",
  "function": "argue_timing",
  "density": "overcrowded",
  "nDataPoints": 15,
  "notes": "The chart shows a trend of AI model performance over time, highlighting models trained with RL (specifically o1-pro and o3) as the top performers.",
  "elementsJson": [
    "headline_text",
    "action_title",
    "line_chart",
    "paragraph",
    "callout_box"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5fec-763b-95d0-451a96d1edee/2",
  "deckHref": "/decks/019dd923-5fec-763b-95d0-451a96d1edee",
  "deckJsonHref": "/decks/019dd923-5fec-763b-95d0-451a96d1edee.json",
  "deckAnchorHref": "/decks/019dd923-5fec-763b-95d0-451a96d1edee#slide-2",
  "components": [
    {
      "bbox": {
        "h": 0.15,
        "w": 0.72,
        "x": 0.14,
        "y": 0.1
      },
      "kind": "callout",
      "text": "Reinforcement learning is the only way for AI models to acquire reasoning, planning and true agency",
      "attrs": null,
      "subkind": "primary",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "1237383c-0c21-4251-8d9e-d15d8f08d836",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.1,
        "w": 0.72,
        "x": 0.14,
        "y": 0.85
      },
      "kind": "callout",
      "text": "However, very few companies have the expertise, time or money to leverage this powerful technology.",
      "attrs": null,
      "subkind": "primary",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "dd27dcdf-2abe-4648-86a1-a30b15319165",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "callout",
      "text": "Modern LLMs can only achieve advanced reasoning capabilities through the use of reinforcement learning.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-bb81-722c-82e0-7910775579a3",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.56,
        "x": 0.04,
        "y": 0.315
      },
      "kind": "chart",
      "text": "Model Performance: GPQA Score over time",
      "attrs": null,
      "subkind": "line",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f17ec422-9631-46d2-bd8f-3bd677fa6864",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "GPQA Score",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-bb81-722c-82e0-7dbccaba833d",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.33,
        "x": 0.63,
        "y": 0.32
      },
      "kind": "paragraph",
      "text": "Beyond human knowledge. To achieve superhuman performance, AI models cannot rely solely on human-created examples. Reinforcement learning allows models to go beyond that by learning through trial and error. Modern LLMs can only achieve advanced reasoning capabilities through the use of reinforcement learning.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "5b94e8a1-3294-4c3d-a225-25a4c01c2ddc",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.03,
        "w": 0.15,
        "x": 0.02,
        "y": 0.03
      },
      "kind": "title",
      "text": "Reinforcement learning",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "799f5d50-f5cb-4418-b189-70af2eb4c4e5",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "AIDA Model",
      "slug": "aida-model",
      "agent": "Storyteller",
      "layer": "slide",
      "matchId": "bdb12a6d-1b1f-456c-940f-0fa6d0647b30",
      "evidence": "The slide presents a clear problem (achieving superhuman performance in AI models), a solution (reinforcement learning), and a call to action (leveraging this technology).",
      "confidence": 0.7
    }
  ],
  "frameworks": [
    {
      "name": "argument-from-analogy",
      "slug": null,
      "matchId": "196a311f-50d8-41e2-bf09-4684e7b3441d",
      "evidence": "Uses the trend of RL-trained models to argue for the necessity of RL in future AI development.",
      "confidence": 0.8
    }
  ],
  "arcBeats": [],
  "loops": [
    {
      "to": 2,
      "from": 2,
      "name": "Why Now",
      "slug": "15-why-now",
      "bestFor": "Sales pitches, fundraising, requesting immediate budget approval",
      "matchId": "f90ebde1-64a3-41bf-8c26-7e7733f10c61",
      "evidence": "The 'why now' section on page 2 sets up the urgency for investment",
      "position": 0,
      "objective": "Explain why now is the right time for investment",
      "structure": "The Context (Trends) -> The Trigger Event -> The Window of Opportunity",
      "confidence": 0.8,
      "description": "Create temporal urgency by proving that the window of opportunity is opening or closing"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}