{
  "docId": "019dd923-5e88-73ef-bd5d-0d0f98caffe1",
  "docSlug": "5df6bd1b0447b5f6",
  "documentTitle": "2025 Air Street Capital The State of AI Report 2025",
  "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": 313,
  "prevPage": 57,
  "nextPage": 59,
  "slideType": "industry_trends",
  "function": "summarize",
  "density": "overcrowded",
  "nDataPoints": 8,
  "notes": "The slide highlights the technical architecture (MoLE) and the scale of training data (OMat24, OMat25, OMol25).",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "paragraph",
    "line_chart"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-0d0f98caffe1/58",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-0d0f98caffe1#slide-58",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "By replacing DFT with fast and accurate AI surrogates, UMA makes it possible to simulate materials, molecules, and sorbents at a scale that was previously unimaginable.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-41ac-7180-a81f-e56b56a923ff",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.4,
        "w": 0.3,
        "x": 0.68,
        "y": 0.45
      },
      "kind": "chart",
      "text": "Performance and scaling charts (a, b, c, d)",
      "attrs": null,
      "subkind": "line",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "6d33062c-f5f7-4c22-8082-e34339743ac6",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.18,
        "w": 0.65,
        "x": 0.02,
        "y": 0.42
      },
      "kind": "list",
      "text": "UMA is a Mixture of Linear Experts (MoLE) with the largest model reaching 3.7B. It's trained on DFT calculations of 500M atomistic configurations from OMat24 (118M structures, 400M CPU core-hours), OMat25 (88M structures, 600M CPU core-hours), OMol25 (100M molecules, 6B CPU core-hours), and adsorption datasets.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "abf73588-af01-4f70-84ef-3ffd61915e3a",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.65,
        "x": 0.02,
        "y": 0.62
      },
      "kind": "list",
      "text": "UMA goes beyond prior models by embedding charge, spin, and task identity, and by ensuring energy-conserving force predictions for long molecular dynamics rollouts.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "b482f30b-c209-4570-8bdc-8fbc3b65f539",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.65,
        "x": 0.02,
        "y": 0.72
      },
      "kind": "list",
      "text": "Across crystal stability (Matbench Discovery), catalysis (OC20), molecules (OMol25), and sorbents (ODAC25), UMA consistently sets the new standard.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "b931b47c-a8c7-456f-94c9-09e78c016868",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "Val Loss: 3.7B",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-41ac-7180-a81f-e93fe73eee57",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.18,
        "w": 0.65,
        "x": 0.02,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "Meta's FAIR trained UMA, a new family of universal interatomic potentials. These approximate the forces and energies between atoms, a task that usually demands resource-intensive quantum calculations (DFT). By replacing DFT with fast and accurate AI surrogates, UMA makes it possible to simulate materials, molecules, and sorbents at a scale that was previously unimaginable. They also produced the largest materials database.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "3cd1ff74-f8b1-420d-8eb1-5fe133ef9bf7",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.04,
        "w": 0.4,
        "x": 0.02,
        "y": 0.14
      },
      "kind": "title",
      "text": "Scaling to Universal Atomistic Models",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "c65b62ab-f980-4784-8168-768af1c76c03",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 189,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-9f380673831f",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Sections 1-2 lay out research findings and industry data with charts and case studies.",
      "position": 1,
      "confidence": 70,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 89,
      "from": 12,
      "beatId": "019dd95a-0682-776c-8e35-ac68557d958b",
      "arcName": "The Onion",
      "arcSlug": "onion",
      "beatName": "First Layer",
      "beatSlug": "onion-first-layer",
      "evidence": "Research section unpacks technical layer beneath the headlines.",
      "position": 2,
      "confidence": 45,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 67,
      "from": 55,
      "name": "Zoom In",
      "slug": "06-zoom-in",
      "bestFor": "Technical deep-dives, case studies, detailed analysis",
      "matchId": "019dd95a-07fe-70ce-8d3e-9d32ca243ef1",
      "evidence": "Sequence of case studies: AlphaEvolve, UMA, MatterGen, AMIE, AlphaFold reproductions all illustrate AI-as-scientist.",
      "position": 5,
      "objective": "Survey AI's rise as scientific collaborator across biology/chemistry/medicine",
      "structure": "The Big Picture -> Key Area of Focus -> Specific Detail -> Implication",
      "confidence": 75,
      "description": "Start broad, then progressively focus on specific details that prove your point"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}