Emergence of general intelligence through skills

A General Robotics is building general-purpose intelligence for every robot in any form, across any real-world scenario. True generalist robots need more than just a foundation model. They must adapt across morphologies, be simple to program, and meet rigorous safety standards.

We believe general intelligence emerges from rich composition of robot skills — not just larger models. By combining modular AI skills, we enable robots to sense, reason, and act with precision. Our approach is data-efficient, interpretable, and built for safety-critical use.

Making robots useful fast
We’re scientists, engineers, and builders behind some of the most widely adopted robotics and AI tools — from AirSim , Neural Simulation and VideoPoet to open-source projects like Tensorflow Object Detection, ChatGPT for Robotics, ClimaX, and PACT.

Foundations in research.
Our research is grounded in deep learning and safety. We believe that general-purpose intelligence for robots won’t come from monolithic models, but from modular deep learning structures. Real-world deployment demands systems that are composable — modular, adaptable, and inspectable by design. Safety is not a layer to add later; it’s a foundation. That’s why we build mechanisms for blame assignment and interpretability into the architecture itself, ensuring accountable and reliable behavior from the ground up.