Deep Learning 2.0 for mobile robots on earth and in space.




    How to reduce training data demand to a millionth and achieve a hundredfold performance for deep learning deployed on mobile robots under real-world constraints?



    At Demiurge, we are redesigning the units of deep neural networks to solve the fundamental challenges of training data shortage and adversarial examples that are tethering all deep learning applications on mobile robots in the physical world.



    By integrating neuroscience with deep learning in their shared domain of mobile agents that generalize all animals and robots, Demiurge is exclusively focused on translating the computational principles of biological neurons into powerful algorithms to enable ultra-safe fully autonomous cars on earth and self-contained exploratory robots in space.