active-inference
(2)
ThoughtForge AI
ThoughtForge AI is a Berkeley-based robotics software company building what it calls Universal Robotic Intelligence: an Active Inference control platform that lets existing industrial robots, drones, and crawlers adapt in real time to messy, unpredictable environments. Its lightweight models run on a single CPU at the edge, claim over 99% accuracy on tasks as designed, and reportedly require roughly 90,000x less training data than conventional deep learning. Founded by veterans of Microsoft's Bonsai AI team, ThoughtForge sells to manufacturing, energy, aerospace, and logistics operators who want to upgrade the robots they already own rather than replace them.
Matthew Brown
Matthew Brown is the founder and CEO of ThoughtForge AI, a Berkeley robotics company building a cloud-based control platform that gives existing industrial robots real-time adaptivity and human-level dexterity using Active Inference, a biologically-inspired approach drawn from neuroscience. Over more than two decades in AI, he wrote combat logic for the Aegis system at Lockheed Martin, built the enemy intelligence behind game franchises like Halo, Bioshock, and Splinter Cell, and was the first engineer at Bonsai AI, where he helped ship one of the world's first production deep reinforcement learning platforms before Microsoft acquired it in 2018. He now teaches and sits on the board of the Active Institute while ThoughtForge works with manufacturers like bp, Siemens, and members of the ARM Institute to bring common sense to the factory floor.