ThoughtForge AI Adaptive Robotic Control

ThoughtForge AI, a Berkeley-based startup, has developed a new approach to robotic control that is accurate and robust with real-time multimodal adaptivity for dexterous and dynamic tasks in real-world environments.

ThoughtForge’s control models’ unique architecture and adaptive capabilities enable models to use 90,000x less data than existing ML techniques to train, deploy on edge using a single CPU, and continuously learn and improve post-deployment. Our highly structured networks are task-specific, work seamlessly with robots' existing sensors, and adapt within 1-3 milliseconds to complete a task with at least 99.7% accuracy, even as tasks deviate or the environment changes.