The Vidar embodied AI mannequin from ShengShu makes use of simulated worlds as a substitute of bodily coaching knowledge. Supply: Adobe Inventory, Vectorhub by ice
ShengShu Know-how Co. yesterday launched its multi-view bodily AI coaching mannequin, Vidar — which stands for for “video diffusion for motion reasoning.” Utilizing Vidu’s capabilities in semantic and video understanding, Vidar makes use of a restricted set of bodily knowledge to simulate a robotic’s decision-making in real-world environments, stated the corporate.
“Vidar presents a radically totally different strategy to coaching embodied AI fashions,” acknowledged ShengShu Know-how. “Simply as Tesla focuses on vision-based coaching and Waymo leans into lidar, the {industry} is exploring divergent paths to bodily AI.”
Based in March 2023, ShengShu Know-how specializes within the improvement of multimodal giant language fashions (LLMs). The Beijing-based firm stated it delivers mobility-as-a-service (MaaS) and software-as-a-service (SaaS) merchandise for smarter, quicker, and extra scalable content material creation.
With its flagship video-generation platform Vidu, ShengShu stated it has reached customers in additional than 200 international locations and areas world wide, spanning fields together with interactive leisure, promoting, movie, animation, cultural tourism, and extra.
Vidar simulated coaching to speed up robotic improvement
“Whereas some firms practice bodily AI by embedding fashions into real-world robots and amassing knowledge by the bodily interactions that their robots encounter, it’s a way that’s pricey, hardware-dependent, and troublesome to scale,” stated ShengShu Know-how. “Others depend on purely simulated coaching, however this typically lacks the variability and edge-case knowledge wanted for real-world deployment.”
Vidar takes a unique strategy, the corporate claimed. It combines restricted bodily coaching knowledge with generative video to make predictions and generate new hypothetical eventualities, making a multi-view simulation that includes lifelike coaching environments, all inside a digital house. This permits for extra sturdy, scalable coaching with out the time, value, or limitations of physical-world knowledge assortment, defined ShengShu.
Constructed on high of the Vidu generative video mannequin, Vidar can carry out dual-arm manipulation duties with multi-view video prediction and even reply to natural-language voice instructions after fine-tuning. The mannequin successfully serves as a digital mind for real-world motion, stated the corporate.
Utilizing Vidu’s generative video engine, Vidar generates large-scale simulations to cut back dependency on bodily knowledge, whereas sustaining the complexity and richness wanted to coach real-world-capable AI brokers. ShengShu stated Vidar can extrapolate a generalized sequence of robotic actions and duties from solely 20 minutes of coaching knowledge. The corporate asserted that’s between 1/80 and 1/1,200 of the info wanted to coach industry-leading fashions together with RDT and π0.5.
ShengShu stated Vidar’s core innovation lies in its modular two-stage studying structure. Not like conventional strategies that merge notion and management, Vidar decouples them into two distinct levels for larger flexibility and scalability.
Within the upstream stage, large-scale basic video knowledge and moderate-scale embodied video knowledge are used to coach Vidu’s mannequin for perceptual understanding.
Within the second downstream stage, a task-agnostic mannequin referred to as AnyPos turns that visible understanding into actionable motor instructions for robots. This separation makes it considerably simpler and quicker to coach and deploy AI throughout several types of robots, whereas reducing prices and rising scalability.
Vidar is designed to cut back the quantity of coaching knowledge wanted to coach AI fashions. Supply: ShengShu Know-how.
Vidar a framework for scalable embodied intelligence
Vidar follows a scalable coaching framework impressed by language and picture basis fashions of the previous decade of AI breakthroughs. ShengShu stated its three-tiered knowledge pyramid, spanning large-scale generic video, embodied video knowledge, and robot-specific examples, makes for a extra versatile system, lowering conventional knowledge bottleneck.
Constructed on the U-ViT structure, which explores the fusion of diffusion fashions and transformer architectures for a large assortment of multimodal technology duties, Vidar harnesses long-term temporal modeling and multi-angle video consistency to energy bodily grounded decision-making.
This design helps speedy switch from simulation to real-world deployment, which ShengShu stated is important for robotics in dynamic environments. It additionally minimizes engineering complexity, in accordance with the corporate,
ShengShu stated Vidar can facilitate robotics adoption throughout a number of sectors. From dwelling assistants and eldercare to good manufacturing and medical robotics, the mannequin permits quick adaptation to new environments and multi-task eventualities, all with minimal knowledge, it added.
Vidar creates an AI-native path for robotics improvement that’s environment friendly, scalable, and cost-effective, ShengShu claimed. By reworking basic video into actionable robotic intelligence, the corporate stated its mannequin can bridge the hole between visible understanding and embodied company.
Vidar has a modular studying structure. Supply: ShengShu Know-how
ShengShu marks milestones in multimodal AI
Vidar builds on the speedy momentum of the Vidu video basis mannequin, stated ShengShu. The corporate listed statistics since its debut:
- Vidu reached 1 million customers inside one month
- Surpassed 10 million customers in simply three months
- Generated over 100 million movies by Month 4
- Reference-to-video technology exceeded 100 million by Month 8
- Complete generated movies now high 300 million
ShengShu continues to increase the frontiers of multimodal AI, Vidar represents the following frontier—bringing generalization, generativity, and embodiment into one unified system.
Editor’s word: RoboBusiness 2025, which will probably be on Oct. 15 and 16 in Santa Clara, Calif., will embrace tracks on bodily AI and humanoid robots. Registration is now open.
