{
  "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": 115,
  "pageCount": 213,
  "prevPage": 114,
  "nextPage": 116,
  "slideType": "industry_trends",
  "function": "analyze_data",
  "density": "dense",
  "nDataPoints": 2,
  "notes": "Includes a technical diagram comparing model architectures (Dense vs MoE vs Hybrid).",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "paragraph",
    "screenshot"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0856e1444fb9/115",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0856e1444fb9#slide-115",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "With readily available open source frontier models, the appeal of training custom models is increasingly unclear.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d1a-ff28a8fdb003",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.28,
        "x": 0.67,
        "y": 0.5
      },
      "kind": "image",
      "text": "Diagram of model architectures: Dense Transformer (Llama 2, 3), MoE Transformer (Mixtral, Grok, DBRX), and Dense-MoE Hybrid Transformer (Snowflake Arctic).",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "aadf5426-4b88-4a39-baf5-e8707b705cab",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.55,
        "w": 0.6,
        "x": 0.05,
        "y": 0.3
      },
      "kind": "list",
      "text": "The Mosaic research team, now folded into Databricks, open-sourced DBRX in March. A 132B MoE model, DBRX was trained on just over 3,000 NVIDIA GPUs at a cost of $10M. Databricks is pitching the model as a foundation for enterprises to build on and customize, while remaining in control of their own data.\nMeanwhile, Snowflake's Arctic is pitched as the most efficient model for enterprise workflows, based on a set of metrics covering tasks including coding and instruction following.\nIt's unclear how much enterprises are willing to invest in costly custom model tuning, given the constant set of releases and improvements driven bigger players.\nWith readily available open source frontier models, the appeal of training custom models is increasingly unclear.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "becf0d49-f35e-4bc6-b908-3c4aefa6d836",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.6,
        "x": 0.05,
        "y": 0.2
      },
      "kind": "paragraph",
      "text": "In last year's report, we touched on Databricks and Mosaic's LLM combined strategy, which focused on fine-tuning models on customer's data. Is the 'bring your own model' era over?",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "27dc0367-b17a-45a8-b2e6-b510fbba1866",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.6,
        "x": 0.05,
        "y": 0.13
      },
      "kind": "title",
      "text": "Databricks and Snowflake pivot to build their own models...but can they compete?",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "e0c66450-8370-4ea9-aa5e-a46375fcecc6",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-1055-74e0-a959-cf257a50fbbb",
      "evidence": "Title 'Databricks and Snowflake pivot to build their own models...but can they compete?'",
      "confidence": 78
    }
  ],
  "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": 174,
      "from": 86,
      "beatId": "019dd95a-0682-776c-8e35-73133c63a096",
      "arcName": "Voyage and Return",
      "arcSlug": "voyage-return",
      "beatName": "Discoveries",
      "beatSlug": "voyage-return-discoveries",
      "evidence": "Industry + Politics findings on NVIDIA, regulation, geopolitics",
      "position": 3,
      "confidence": 55,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 121,
      "from": 112,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3e-53b7db321fbe",
      "evidence": "Chat sidekicks, labs as products, Mistral, Databricks/Snowflake, regulators, GitHub, agents, search",
      "position": 13,
      "objective": "Track AI products & dev-tools ecosystem maturation",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 72,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}