Deployment of 'world model systems' — AI that accurately simulates and anticipates the physical and thermodynamic dynamics of reality — serves as the critical foundational training ground for embodied agents, letting them experience billions of hours o...
Predictor: Demis Hassabis
Prediction text
Deployment of 'world model systems' — AI that accurately simulates and anticipates the physical and thermodynamic dynamics of reality — serves as the critical foundational training ground for embodied agents, letting them experience billions of hours of simulated physical trial-and-error before real-world deployment. | First humanoid deployment trained purely from sim
Key catalyst: First humanoid deployment trained purely from sim
Watch events: Genie 4+ releases; NVIDIA Cosmos adoption
Resolution evidence
DeepMind Genie 2 / Genie 3 world models; NVIDIA Cosmos, Tesla FSD sim; Figure Helix training. Billion-hour simulation scales documented.
Predictor: Demis Hassabis
Calibration plot (stated vs observed)
Evidence about this node from Demis Hassabis is multiplied by κ in /api/intake. Lower κ = less weight; floors at 0.10 (effectively silenced) and caps at 1.00 (full weight).
Reference class: humanoid_commercial_volume
>10,000 unit cumulative deployment of humanoid robot SKU within 3 years of debut
Tetlock-style outside view: at TRF=1 (just predicted), outside view dominates (w_in=0.3). At TRF=0 (deadline), inside view dominates (w_in=1.0). The blend regularizes overconfident inside views toward the historical base rate.
Probability over time
Milestone chain
- 2026-03-18hit1X Technologies trains NEO Gamma humanoid using Cosmos Predict + TransferHow: 1X publicly confirms Cosmos Predict + Cosmos Transfer used to train NEO Gamma humanoid, per NVIDIA partner announcementSource: https://nvidianews.nvidia.com/news/nvidia-and-global-robotics-leaders-take-physical-ai-to-the-real-worldconf 95%Notes: HIT — first concrete instance of WFM-trained humanoid going to production.
- 2026-04-15hitNVIDIA announces Cosmos 3 unifying world generation + vision reasoning + action simHow: NVIDIA publishes Cosmos 3 release notes describing the first world foundation model unifying synthetic world generation, vision reasoning, and action simulationSource: https://www.nvidia.com/en-us/ai/cosmos/ — NVIDIA Cosmos product page describing Cosmos 3 as unified WFMconf 85%Notes: HIT — confirms world-model thesis; foundational training infrastructure for embodied agents.
- 2026-04-15hitNVIDIA Isaac GR00T N1.6 ships with Cosmos Reason for full-body humanoid controlHow: NVIDIA releases Isaac GR00T N1.6 open-reasoning vision-language-action model purpose-built for humanoids, leveraging Cosmos ReasonSource: https://developer.nvidia.com/isaac/gr00tconf 95%
- 2026-10-13pendingQ1 window check-in (25%)
- 2027-07-25pendingQ2 window check-in (50%)
- 2027-01-01 → 2029-02-14pendingFirst commercial humanoid deployment whose pre-training was 100% in simulationHow: Public announcement (Figure / 1X / Apptronik / Sanctuary) of a humanoid deployed in commercial workforce role whose policy was trained entirely in WFM simulation with zero real-world bootstrap fine-tuningSource: Humanoids Daily / company press / NVIDIA partner releasesconf 45%Notes: This is the prediction's named milestone. Sim-to-real zero-shot transfer remains the open frontier.
- 2028-05-05pendingQ3 window check-in (75%)
- 2027-06-01 → 2029-12-31pending≥10,000 cumulative humanoid hours of WFM-trained operation in commercial settingsHow: Public reporting (NVIDIA partner releases or industry trackers) shows aggregate ≥10,000 commercial-deployment hours from WFM-trained humanoid policiesSource: NVIDIA + Humanoids Daily reportingconf 40%Notes: Cascade — validates the 'billions of hours of simulated trial-and-error' thesis with real-world hours.
No downstream cascades — this prediction is a leaf in the dependency graph.
What if this resolves?
Click a button to clamp this prediction and run a Gibbs sample. Returns the predictions whose marginals shift most. ~30s per run; ideal for stress-testing "if X resolves, what else moves?"
Evidence chain
Network propagation neighbors
No propagation data yet. Run inference/.venv/bin/python scripts/ops/run_loopy_belief_propagation.py on the droplet, or wait for the Sunday 02:00 UTC weekly cron.
Prerequisites (1)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| correlate | S_AGI_MID_2029 | AGI mid: Kurzweil 2029 path | agi_general_capability | — |
Dependents (0)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| No dependents | ||||
Expected milestones (1)
| Expected by | Description | Status |
|---|---|---|
| 2030-12-31 | [Robotics 2030-12] [CYB_019] Genie 4+ releases; NVIDIA Cosmos adoption [230_022] Shanghai factory commitment announcement [CMQ_050] Annual humanoid unit shipments; BMW / Mercedes / Figure deployment ramp; Tesla Optimus productio | pending |
Validations (1)
| Observed at | Status | By | Notes |
|---|---|---|---|
| 2026-04-29 | partial | thesis_timeline_v1.0_import | DeepMind Genie 2 / Genie 3 world models; NVIDIA Cosmos, Tesla FSD sim; Figure Helix training. Billion-hour simulation scales documented. |
Linked documents (10)
Raw metadata
{
"nia": false,
"mode": "FORECAST",
"role": "Cited-CEO",
"context": "Distinct from AI_005 (jagged intelligence), AI_031 (100 years of biology). Specific robotics-training framing. Couples with INF_025 (Figure universal brain), 229_042 (Figure omni-model).",
"to_year": 2030,
"conv_cues": "CEO FIRST_PERSON; specific technical framing",
"direction": "HAPPEN",
"from_year": 2026,
"timeframe": "2026-2030",
"conv_level": "HIGH",
"milestones": [
{
"kind": "llm_pre_event",
"label": "1X Technologies trains NEO Gamma humanoid using Cosmos Predict + Transfer",
"notes": "HIT — first concrete instance of WFM-trained humanoid going to production.",
"source": "https://nvidianews.nvidia.com/news/nvidia-and-global-robotics-leaders-take-physical-ai-to-the-real-world",
"status": "hit",
"weight": 0.4,
"ordinal": -8,
"source_id": null,
"confidence": 0.95,
"source_url": "https://nvidianews.nvidia.com/news/nvidia-and-global-robotics-leaders-take-physical-ai-to-the-real-world",
"expected_date": "2026-04-01",
"observed_date": "2026-03-18",
"research_origin": "deep_research",
"expected_date_range": {
"to": "2026-06-30",
"from": "2026-01-01"
},
"measurement_criterion": "1X publicly confirms Cosmos Predict + Cosmos Transfer used to train NEO Gamma humanoid, per NVIDIA partner announcement"
},
{
"kind": "llm_pre_event",
"label": "NVIDIA announces Cosmos 3 unifying world generation + vision reasoning + action sim",
"notes": "HIT — confirms world-model thesis; foundational training infrastructure for embodied agents.",
"source": "https://www.nvidia.com/en-us/ai/cosmos/ — NVIDIA Cosmos product page describing Cosmos 3 as unified WFM",
"status": "hit",
"weight": 0.4,
"ordinal": -7,
"source_id": null,
"confidence": 0.85,
"source_url": "https://www.nvidia.com/en-us/ai/cosmos/",
"expected_date": "2026-04-15",
"observed_date": "2026-04-15",
"research_origin": "deep_research",
"expected_date_range": {
"to": "2026-05-31",
"from": "2026-03-01"
},
"measurement_criterion": "NVIDIA publishes Cosmos 3 release notes describing the first world foundation model unifying synthetic world generation, vision reasoning, and action simulation"
},
{
"kind": "llm_pre_event",
"label": "NVIDIA Isaac GR00T N1.6 ships with Cosmos Reason for full-body humanoid control",
"source": "https://developer.nvidia.com/isaac/gr00t",
"status": "hit",
"weight": 0.4,
"ordinal": -6,
"source_id": null,
"confidence": 0.95,
"source_url": "https://developer.nvidia.com/isaac/gr00t",
"expected_date": "2026-04-30",
"observed_date": "2026-04-15",
"research_origin": "deep_research",
"expected_date_range": {
"to": "2026-06-30",
"from": "2026-03-01"
},
"measurement_criterion": "NVIDIA releases Isaac GR00T N1.6 open-reasoning vision-language-action model purpose-built for humanoids, leveraging Cosmos Reason"
},
{
"kind": "quartile_checkpoint",
"label": "Q1 window check-in (25%)",
"status": "pending",
"weight": 0.05,
"ordinal": -5,
"source_id": null,
"expected_date": "2026-10-13",
"observed_date": null
},
{
"kind": "quartile_checkpoint",
"label": "Q2 window check-in (50%)",
"status": "pending",
"weight": 0.05,
"ordinal": -4,
"source_id": null,
"expected_date": "2027-07-25",
"observed_date": null
},
{
"kind": "llm_pre_event",
"label": "First commercial humanoid deployment whose pre-training was 100% in simulation",
"notes": "This is the prediction's named milestone. Sim-to-real zero-shot transfer remains the open frontier.",
"source": "Humanoids Daily / company press / NVIDIA partner releases",
"status
... (truncated)