Understand this story
In an interview, TorqueAGI founder Ashutosh Saxena said that most robot AI is just language or vision models with robot data added. He said TorqueAGI instead uses spatiotemporal graphs to build a World-Reasoning Model, incorporating physics and geometry into reasoning, so new tasks typically require only a few dozen examples rather than tens of thousands. The model has already run on single-arm and dual-arm robots, mobile platforms, humanoids, and tractors, with body-specific details needing to be described for each machine. TorqueAGI is an Nvidia Perception Partner and was selected for the John Deere 2026 Startup Collaborator Program.
AI-generated from public reports
Latest development · Report
Interview with TorqueAGI Founder: Most Robot AI Is Just a Language or Vision Model with Robot Data AddedDevelopments
1 confirmed developments
Saxena says most robotics AI is just language or vision models with robotics data added
Source:Robotics & Automation News
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 2
Robotics & Automation News
Interview with TorqueAGI Founder: Most Robot AI Is Just a Language or Vision Model with Robot Data Added
TorqueAGI founder Ashutosh Saxena says most robot AI is just a language or vision model with robot data added, whereas TorqueAGI uses spatiotemporal graphs to build a World-Reasoning Model that incorporates physics and geometry into reasoning, and new tasks typically require only a few dozen examples rather than tens of thousands. The model has already run on single-arm and dual-arm robots, mobile platforms, humanoids, and tractors, and body-specific details must be described machine by machine. TorqueAGI is an Nvidia Perception Partner and was selected for the John Deere 2026 Startup Collaborator Program.