{
  "docId": "019dd923-5e88-73ef-bd5d-06b04d219fea",
  "docSlug": "dd91c78f6570bf29",
  "documentTitle": "2023 Air Street Capital The State of AI Report 2023",
  "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": 58,
  "pageCount": 163,
  "prevPage": 57,
  "nextPage": 59,
  "slideType": "industry_trends",
  "function": "summarize",
  "density": "overcrowded",
  "nDataPoints": 35,
  "notes": "Includes a spectrogram image and a performance comparison table for various music generation models.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "data_table",
    "photo"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-06b04d219fea/58",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea#slide-58",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "New models from Google, Meta, and the open source community significantly advance the quality of controllable music generation.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a225-b409fb1bdf3b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.15,
        "x": 0.17,
        "y": 0.7
      },
      "kind": "image",
      "text": "funk bassline with a jazzy saxophone solo",
      "attrs": null,
      "subkind": "illustration",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "37fe5abc-2ec1-45a7-b47f-281e2e2528ad",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.1,
        "w": 0.9,
        "x": 0.03,
        "y": 0.31
      },
      "kind": "list",
      "text": "Though not the best in terms of generated music quality, Riffusion was probably the most innovative model. Researchers fine-tuned Stable Diffusion on images of spectrograms, which are then converted into audio clips.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "3f9aaa7b-6619-4c4d-9eba-5b6f635f0df2",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.9,
        "x": 0.03,
        "y": 0.52
      },
      "kind": "list",
      "text": "To our ears, Meta’s MusicGen strikes a better balance between adhering to text descriptions and generating a pleasant melody. It uses a single transformer LM and careful codebook interleaving techniques.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "66fed3a8-22c4-46c9-89f7-64e9f086d29b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.9,
        "x": 0.03,
        "y": 0.42
      },
      "kind": "list",
      "text": "With MusicLM, Google researchers “cast conditional music generation as a hierarchical seq2seq modeling task”. They are able to generate consistent music (@24kHz) over several minutes.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "72b55bd4-a972-4bd4-b1af-4387faedd573",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "MusicCaps Test Set performance",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a225-b9dc84648271",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.07,
        "w": 0.9,
        "x": 0.03,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "New models from Google, Meta, and the open source community significantly advance the quality of controllable music generation.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "289ac4dc-3f60-4a5a-bf8d-4724d22be0c0",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.6,
        "x": 0.34,
        "y": 0.7
      },
      "kind": "table",
      "text": "Performance metrics for Riffusion, Mousai, MusicLM, Noise2Music, and MusicGen variants across FAD, KL, CLAP, OVL, and REL metrics.",
      "attrs": null,
      "subkind": "data",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "18dcf391-0881-49d5-ab14-a8a55aa48bd5",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.04,
        "w": 0.45,
        "x": 0.03,
        "y": 0.14
      },
      "kind": "title",
      "text": "Another year of progress in music generation",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "d9bbb2b8-43ae-48f4-afdf-948f941d151d",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Table data",
      "slug": "table-data",
      "agent": null,
      "layer": "slide",
      "matchId": "214a9c6b-7429-45d9-9043-fdc443becd11",
      "evidence": "table/data: Performance metrics for Riffusion, Mousai, MusicLM, Noise2Music, and MusicGen variants across FAD, KL, CLAP, OVL, and REL metrics.",
      "confidence": 0.8
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 120,
      "from": 11,
      "beatId": "019dd95a-0682-776c-8e35-41afd44ef59f",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Research + Industry sections inventory model, compute, funding facts.",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 120,
      "from": 11,
      "beatId": "019dd95a-0682-776c-8e35-523bfb7f96e6",
      "arcName": "The Mountain",
      "arcSlug": "mountain",
      "beatName": "Rising Action",
      "beatSlug": "mountain-rising-action",
      "evidence": "Escalating capabilities, compute concentration and capital flows.",
      "position": 2,
      "confidence": 45,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 58,
      "from": 50,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-f70f6435b722",
      "evidence": "Text-to-video, instruction editing, Gaussian splatting, NeRFs, depth, SAM, DINOv2, weather, music.",
      "position": 6,
      "objective": "Generative media and visual foundation models proliferate",
      "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
}