{
  "docId": "019dd923-5e88-73ef-bd5d-8f9067e49fd6",
  "docSlug": "67108284e9fcdd32",
  "documentTitle": "2024 Benedict Evans 2024 AI eats the world",
  "authorId": "BenedictEvans",
  "authorName": "Benedict Evans",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "independent_analyst",
  "sourceTypeLabel": "Independent analyst",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.778,
  "pageNumber": 71,
  "pageCount": 87,
  "prevPage": 70,
  "nextPage": 72,
  "slideType": "diagnosis",
  "function": "diagnose",
  "density": "dense",
  "nDataPoints": 24,
  "notes": "The chart is Figure 4 from an OpenAI report, showing GPT performance on exams.",
  "elementsJson": [
    "headline_text",
    "paragraph",
    "bar_chart_stacked",
    "footnote"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-8f9067e49fd6/71",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-8f9067e49fd6",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-8f9067e49fd6.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-8f9067e49fd6#slide-71",
  "components": [
    {
      "bbox": {
        "h": 0.7,
        "w": 0.4,
        "x": 0.55,
        "y": 0.1
      },
      "kind": "chart",
      "text": "Exam results (ordered by GPT-3.5 performance)",
      "attrs": null,
      "subkind": "bar-stacked",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f1b249e2-05d0-4d16-a510-d6452910f301",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.4,
        "x": 0.55,
        "y": 0.75
      },
      "kind": "disclaimer",
      "text": "Figure 4. GPT performance on academic and professional exams. In each case, we simulate the conditions and scoring of the real exam.",
      "attrs": null,
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "28d07b6f-eaa2-44c2-bc26-218496cabf96",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Estimated percentile",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-4e51-732b-a108-be5c79b28f54",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.3,
        "w": 0.4,
        "x": 0.05,
        "y": 0.35
      },
      "kind": "paragraph",
      "text": "Is this intelligence? Reasoning? Or information retrieval and pattern matching? Or - the ability to look like people? Should we be benchmarking against things that people are bad at? Or against things that people are good at?",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "13e3c0da-a68f-4304-935d-9569281cbea7",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.1,
        "x": 0.05,
        "y": 0.9
      },
      "kind": "source-note",
      "text": "Source: OpenAI",
      "attrs": null,
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "90a80ce3-f901-4df4-8b13-2fc554d30bf4",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.4,
        "x": 0.05,
        "y": 0.25
      },
      "kind": "title",
      "text": "What are we measuring?",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "a2ad3b45-5680-4161-8267-b23655e19e3a",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 72,
      "from": 62,
      "beatId": "019dd95a-07a7-7579-bd65-36e6c9c1eb5f",
      "arcName": "The Onion",
      "arcSlug": "onion",
      "beatName": "Core Insight",
      "beatSlug": "onion-core-insight",
      "evidence": "Section 'A path to AGI?' — three views of understanding, 'we don't know'.",
      "position": 4,
      "confidence": 82,
      "parentBeatName": "Turn",
      "parentBeatSlug": "turn"
    },
    {
      "to": 72,
      "from": 14,
      "beatId": "019dd95a-07a7-7579-bd65-41b6cdc28844",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Implications (So What)",
      "beatSlug": "triple-take-the-implications-so-what",
      "evidence": "Platform shift, value capture, agents, AGI debate.",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Reflection",
      "parentBeatSlug": "reflection"
    }
  ],
  "loops": [
    {
      "to": 72,
      "from": 63,
      "name": "Hypothesis Test",
      "slug": "42-hypothesis-test",
      "bestFor": "Strategy validation, due diligence, research presentations",
      "matchId": "019dd95a-08f8-7619-ac2e-b4b318b6911b",
      "evidence": "Three views slide + 'we don't have a chart for AGI' resolves to inconclusive.",
      "position": 7,
      "objective": "Test competing hypotheses on whether LLMs can reach AGI-level understanding",
      "structure": "The Hypothesis -> Test 1 (Supports/Refutes) -> Test 2 (Supports/Refutes) -> Conclusion (Validated/Pivoted)",
      "confidence": 75,
      "description": "State a hypothesis upfront, then systematically prove or disprove it with evidence"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}