Shanghai-based Wiz Technology has unveiled its pioneering Fidget spatiotemporal foundation model, a significant advancement in urban artificial intelligence. The launch event in Zhangjiang also introduced the Spatiotemporal Metaverse, an urban agent system designed for city forecasting and automated operations. This new technology aims to transform how cities are managed by predicting their future states and enabling autonomous responses.
A New Paradigm in Urban AI
Spatiotemporal foundation models represent a new class of AI trained on diverse space-time data for multifaceted forecasting and planning tasks. Unlike general language models that predict the next word, this technology forecasts a city's "next state," according to Wiz Technology CEO Tao Chuang. The Fidget model's core capabilities include advanced urban modeling, prediction, memory, and continuous learning.
The system functions by integrating a city's complex spatial structure with its temporal evolution into a unified digital representation. This is achieved through sophisticated spatial grid encoding and the generation of spatiotemporal token vectors. This foundational approach allows the AI to understand and interpret the intricate dynamics of an urban environment comprehensively.
Training and Performance in Zhangjiang
The model was rigorously trained in the Zhangjiang AI Innovation Town, utilizing a massive dataset of over 20,000 categories of multimodal information. This data included everything from geographic layouts and enterprise activity to pedestrian flow, transportation patterns, and camera feeds. In real-world tests within the town, the model demonstrated remarkable accuracy, achieving 95% in transportation and 97% in parking predictions.
The "Spatiotemporal Metaverse" in Action
A live demonstration showcased the system's practical capabilities in a fully autonomous, closed-loop operation. Based on pedestrian and weather forecasts, the system dispatched a drone 1.5 kilometers to identify accumulated garbage on a road. After returning images for verification, the AI automatically generated a cleanup work order and deployed an unmanned cleaning vehicle to the site.
An Integrated Urban Intelligence System
CEO Tao Chuang described the Spatiotemporal Metaverse as giving a city its own "senses," "brain," and "limbs." The city's "senses" are its active perception systems, including cameras and sensors that gather real-time data. This information is then processed by the "brain," a cluster of digital agents that predict future states for traffic, pedestrian flow, and more.
The system's "limbs" are a fleet of physical agents, such as drones, unmanned vehicles, and robots. These agents execute the decisions made by the digital brain, translating AI-driven analysis into tangible actions in the physical world. This integration of digital intelligence with physical automation paves the way for truly autonomous urban operations.
Future Implications and Collaborative Development
The project's success is built on a collaborative framework guided by the Shanghai Pudong New Area Data Bureau. This model combines government data, real-world scenarios from the town, and technology from Wiz and its ecosystem partners. The system is designed to improve continuously through a "data flywheel," where every action provides feedback that refines the model's understanding.
The launch of the Fidget model and Spatiotemporal Metaverse marks a pivotal moment for smart city development. Wiz Technology has created a powerful tool that moves beyond simple monitoring to enable predictive and autonomous urban management. This innovation establishes a new paradigm for intelligent governance, promising more efficient, responsive, and self-regulating cities in the future.