True AGI requires 'physical intelligence' — an AGI system must be able to control a robotic chassis to play sports or perform complex physical manipulations at amazing levels to truly mirror human brain architecture; software alone cannot achieve full ...

True AGI requires 'physical intelligence' — an AGI system must be able to control a robotic chassis to play sports or perform complex physical manipulations at amazing levels to truly mirror human brain architecture; software alone cannot achieve full general intelligence without embodied interactions. | First published AGI-benchmark definition including embodiment

Key catalyst: First published AGI-benchmark definition including embodiment

Watch events: DeepMind Gemini Robotics next release; humanoid-robot dexterity benchmarks

DeepMind Gemini Robotics / DeepMind GNoME / Figure Helix all embody this thesis; jury out on whether physical integration is strictly necessary for AGI.

{
  "nia": false,
  "mode": "FORECAST",
  "role": "Cited-CEO",
  "context": "Hassabis bridge thesis: digital AGI cannot stand alone; embodiment is prerequisite. Couples with ROB_006 (Huang $50T Physical AI), ROB_009 (Musk 10B robots).",
  "to_year": 2036,
  "conv_cues": "CEO FIRST_PERSON; AGI-definition framing",
  "direction": "HAPPEN",
  "from_year": 2027,
  "timeframe": "2027-2036",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "Vision-language-action (VLA) foundation model demonstrates zero-shot transfer across >=1000 manipulation tasks",
      "source": "Helix (Figure), GR00T (NVIDIA), Pi0 (Physical Intelligence)",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -10,
      "source_id": null,
      "confidence": 0.6,
      "expected_date": "2027-09-30",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2028-12-31",
        "from": "2026-06-30"
      },
      "measurement_criterion": "Published benchmark (RoboArena, ManiSkill, Open X-Embodiment) showing single VLA model handling >=1000 distinct manipulation tasks at >=80% success without per-task finetuning"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -9,
      "source_id": null,
      "expected_date": "2028-09-13",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "Frontier AI lab publicly states physical embodiment is required for AGI claims",
      "source": "Industry debate Hassabis vs. Altman embodiment requirement",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -8,
      "source_id": null,
      "confidence": 0.4,
      "expected_date": "2028-09-30",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2030-06-30",
        "from": "2027-01-01"
      },
      "measurement_criterion": "OpenAI, Anthropic, Google DeepMind, or peer publishes formal position paper or AGI definition that requires demonstrated physical-world capability (manipulation/sport/lab work) as gating criterion"
    },
    {
      "kind": "scenario_signal",
      "label": "Scenario fires: Humanoid R2: 100K+ enterprise by Nov 2028",
      "status": "pending",
      "weight": 0.5,
      "ordinal": -7,
      "source_id": "S_HUMANOID_ENTERPRISE_2028",
      "expected_date": "2028-11-30",
      "observed_date": null
    },
    {
      "kind": "scenario_signal",
      "label": "Scenario fires: AGI mid: Kurzweil 2029 path",
      "status": "pending",
      "weight": 0.5,
      "ordinal": -6,
      "source_id": "S_AGI_MID_2029",
      "expected_date": "2029-03-31",
      "observed_date": null
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "pending",
      "weight": 0.05,
      "ordinal": -5,
      "source_id": null,
      "expected_date": "2030-05-27",
      "observed_date": null
    },
    {
      "kind": "llm_pre_event",
      "label": "Humanoid robot achieves Olympic-level performance on at least one athletic discipline (sub-3-min mile, gymnastics floor routine, etc.)",
      "source": "Unitree H2/G1 sport demos; Figure 03 Helix",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -4,
      "source_id": null,
      "confidence": 0.35,
      "expected_date": "2030-07-02",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2032-12-31",
        "from": "2028-01-01"
      },
      "measurement_criterion": "Verifiable third-party demonstration (broadcast, peer review) of Optimus/Figure/Unitree successor completing an athletic event at performance equivalent to 90th-percentile human Olympic athlete"
    },
    {
      "kind": "llm_pre_event",
      "label": "Tesla Optimus or Figure deploys >=1M units globally",
      "source": "Tesla Gen-2 Gigafactory Texas 10M/yr target 2027",
      "status": "pending",
      "we
... (truncated)