Robotics
Will AI Agents Control Physical Robots?
Don't just read what happened. See what could happen next.
Prediction
Language-and-vision agents increasingly command robots in structured settings within 24 months; general household control remains unreliable.
- Confidence
- 53%
- Horizon
- 12–36 months
- Impact
- High
- Direction
- Increasing
Prediction changed +9 points in 21 days
Short answer
Software agents meeting hardware is the past year’s “physical AI” story. Demos impress; household success rates stay low. Factories and labs will see agent-directed robots before homes do.
Why this question matters
Agent-to-robot control could unlock humanoid usefulness — or expose new safety and liability surfaces.
What's happening now?
Vision-language-action models
Models map instructions to motor commands more generally.
Signal · strong
Sim-to-real gap
High sim scores still fail messy physical spaces.
Signal · strong
Industrial pilot integrations
Agents schedule and supervise robot cells before full autonomy.
Signal · moderate
Safety certification lag
Physical harm risk keeps humans nearby.
Signal · moderate
What could happen next?
Scenario A
Factory agent control
Agents become the UI for fleets of industrial robots.
Scenario B
Teleop + assist
Humans remain in the loop; agents only suggest motions.
Scenario C
Home breakthrough delayed
Household robotics stays novelty despite agent hype.
Key companies / entities
- Figure
- Tesla
- Google DeepMind
- Covariant
- Boston Dynamics
Evidence
- research
VLA and robot foundation-model papers
- research
Stanford AI Index robotics stats
- company
Humanoid + AI control demos
Prediction history
- Sep 22, 202653%
- Sep 15, 202650%
- Sep 8, 202647%
- Sep 1, 202644%