NVIDIA Quietly Dropped a Robot World Model That Nobody Expected
Автор: AISoftScope
Загружено: 2026-02-25
Просмотров: 11
Описание:
NVIDIA just released DreamDojo, a fully open-source robot world model trained on 44,711 hours of real human video — and it changes how we teach robots to understand the physical world. Instead of manually coding physics rules, DreamDojo simply watches humans go about their daily lives and learns how reality works from that. The result is a simulator that matches real-world robot performance with a 0.995 correlation score, improves success rates by 17% through AI planning, and runs fast enough for real-time use at 10.81 frames per second.
In this video we break down exactly how DreamDojo works, what makes the dataset so special, how NVIDIA solved the problem of turning human video into robot instructions, and what the real-world performance numbers actually mean for the future of robotics.
🔗 Paper: https://arxiv.org/pdf/2602.06949
🔗 GitHub: https://github.com/NVIDIA/DreamDojo
If you found this useful, subscribe for daily breakdowns of the most important AI and tech releases. New video every day.
#NVIDIA #Robotics #ArtificialIntelligence #DreamDojo #RoboticsAI #MachineLearning #OpenSource #AINews
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
