{
  "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": 64,
  "pageCount": 163,
  "prevPage": 63,
  "nextPage": 65,
  "slideType": "industry_trends",
  "function": "present_solution",
  "density": "dense",
  "nDataPoints": 1,
  "notes": "The slide discusses MultiMedBench (14 tasks) and the architectural shift to multimodal medical models.",
  "elementsJson": [
    "headline_text",
    "paragraph",
    "process_diagram",
    "flowchart",
    "logo_grid"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-06b04d219fea/64",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-06b04d219fea#slide-64",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "To bridge beyond text-based medical Q&A, Google first created MultiMedBench - a 14 task dataset that includes medical Q&A, mammography and dermatology image interpretation, radiology report generation and summarization, and genomic variant calling.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c4-7719-a226-3a910a46198e",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.45,
        "x": 0.52,
        "y": 0.61
      },
      "kind": "diagram",
      "text": "ELIXR architecture diagram showing Image Encoder, Adapter, and LLM",
      "attrs": null,
      "subkind": "process",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "c243aa29-5676-44d3-a28f-a3740936b66d",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.45,
        "x": 0.02,
        "y": 0.58
      },
      "kind": "image",
      "text": "Med-PaLM M architecture and MultiMedBench diagram",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "0486fb39-4d94-4f7c-985a-d4776366f1cd",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.35,
        "w": 0.96,
        "x": 0.02,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "To bridge beyond text-based medical Q&A, Google first created MultiMedBench - a 14 task dataset that includes medical Q&A, mammography and dermatology image interpretation, radiology report generation and summarization, and genomic variant calling. This dataset is used to train a large single multitask, multimodal version of MedPaLM with the same set of model weights. The system exhibits novel emergent capabilities such as generalisation to novel medical concepts and tasks. An alternative lighter-weight approach, ELIXR, was also proposed. ELIXR grafts language-aligned vision encoders onto a fixed LLM, which requires less compute to train and shows promise across tasks including visual QA, semantic search, and zero-shot classification.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "0634d488-9c80-4a2e-92ab-f9a1f04c6f5a",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.04,
        "w": 0.35,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "Next, Med-PaLM goes multimodal",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "ccfe9546-3268-4c83-a926-c3f79c96c2f4",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Canvas framework",
      "slug": "canvas-framework",
      "agent": null,
      "layer": "slide",
      "matchId": "de362575-24ec-4ff0-91fd-b3763002d16a",
      "evidence": "ELIXR architecture diagram showing Image Encoder, Adapter, and LLM",
      "confidence": 0.5
    }
  ],
  "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": 68,
      "from": 59,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-f9a066ea1e78",
      "evidence": "RFdiffusion, ESMFold, PerturbNet, AlphaMissense, Med-PaLM, PLIP, CoDoC, then meta map of AI-for-science fields.",
      "position": 7,
      "objective": "AI for science accumulates breakthroughs across bio and medicine",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 78,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}