appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Coco Feng

A Chinese physical AI start-up has launched a new world model designed to simulate reality by embedding the laws of physics directly into its code – a departure from the data-driven approaches favoured by American tech giants like OpenAI and Meta Platforms.
Shanghai-based Fysics AI announced the launch of the Fysiverse, which it described as a “new-generation physics-based world model that adheres to real-world physical laws”, in a post on its WeChat social media account on Wednesday.
The start-up, founded by former Nvidia senior manager Zhang Lihua, said the model “represents a new paradigm” that could effectively address issues commonly faced by currently available world models, such as “physical illusions, reasoning failures, and breakdowns in non-standard scenarios”.
The world model sector, used to create content and train robots or self-driving technology, is currently dominated by three major paradigms.
The first is video-based generation that duplicates movements by learning from massive video clips. OpenAI’s Sora is a typical example, with the firm describing scaling video generation models as “a promising path towards building general purpose simulators of the physical world”.
The second way is to let the model, without knowledge of physics, construct its own rules of the world in a black box. Meta’s V-JEPA series, short for video joint embedding predictive architecture, is one such system that takes a “self-supervised learning approach”, according to Meta’s website.
The final paradigm relies on three-dimensional modelling to learn, predict and simulate the geometry and dynamics of real environments from text, images, video or coarse 3D elements.
Marble from World Labs, a spatial intelligence company co-founded by Stanford University professor Li Fei-fei, is a typical example.


However, they all have their shortcomings. Fysics AI said the video method only studied correlation, not causation, the knowledge generated in the black box was “confined in a latent space” that was hard to validate, and while 3D models offered strong spatial structure, that did not mean they could handle real-world material, contact and dynamic properties.
It said the Fysiverse was powered by a simulator based on real-world physics knowledge and a video renderer, and the goal was not to generate a vivid video, but to “build a dynamic world that evolves according to the physical laws and is computable, controllable and optimisable”.
It first reconstructs a scene’s 3D geometry by separating foreground objects from the background context, and calculates their movements and interactions using the physics simulator.
Finally, the model renders a lifelike world with “structured information such as geometry, motion, occlusion and contact”, it said.
The company showcased several results, including one in which four cabinets fell like dominoes, and another about a teddy bear bouncing upwards.
Compared with nine unnamed rivals, Fysiverse produced the most natural and realistic scenes, according to the start-up’s demo.
Fysics AI was established in 2024 by Zhang, who was a key contributor to Nvidia’s PhysX simulation engine. He now serves as a vice-dean of the College of Intelligent Robotics and Advanced Manufacturing at Fudan University in Shanghai.
The firm received funding from several institutional investors, including MPC, formerly the China unit of US venture capital firm Matrix Partners, and MetaX, a Chinese designer of graphics processing units.


Philippine tribe vows to stand ground against US-led Pax Silica: ‘listen to us’

As the US weighs restrictions on cloud computing, how will China’s AI sector adapt?

Chinese AI chipmaker Hygon plots expansion from data centres to robotics

For China’s hi-tech IPOs, it’s Hong Kong vs the mainland – but how much does it matter?

Artificial intelligence (AI)

Chinese physical AI start-up proposes new paradigm that bypasses OpenAI, Meta road maps

The logo of Shanghai-based Fysics AI. Photo: Handout

Humanoid robots and fashion models perform on stage during Galaxy Corporation’s Mach33: Physical AI Fashion Show at Galaxy Robot Park in Seoul, South Korea, last month.  Photo: EPA

Alibaba eyes physical world with its first suite of AI models for robots

The Qwen and Alibaba logos are seen in this illustration taken January 29, 2025. Photo: Reuters

The suite splits robot intelligence into three interconnected layers. Photo: Handout

Chinese tech giants, AI ‘godmother’ Li Fei-Fei race to seize the edge in world models

A screenshot of Happy Oyster’s world model. 

Happy Oyster supports two modes of virtual world creation: a directing mode for building a world based on text and image prompts and a wandering mode for exploring that world. 

Li Fei-Fei has unveiled Spark 2.0, an open-source 3D Gaussian splatting rendering engine that aims to give even less powerful devices, such as smartphones, the ability to view large-scale and detailed 3D images. Photo: Handout

Alibaba cuts Qwen AI model costs on Qoder coding platform to capture US workday demand

The Qwen and Alibaba logos. Photo: Reuters

The Alibaba logo on a building in Beijing. Photo: Kyodo

Banks are rushing to recruit AI leadership – but how will the tech be implemented?

Shaown Nandi, the vice-president of technology at Amazon Web Services. Photo: Handout

Customers use HSBC ATMs in Hong Kong’s Central district. Photo: Felix Wong

America’s ‘biggest risk’ on AI is China getting ahead, Bessent says

Treasury Secretary Scott Bessent emphasised that the US had to stay ahead on AI technology. Photo: Reuters

Tech

Tech Trends

Big Tech

Tech War

News

China Future Tech

Shanghai-based Fysics AI says Fysiverse is a ‘new-generation physics-based world model that adheres to real-world physical laws’


Shanghai-based Fysics AI says Fysiverse is a ‘new-generation physics-based world model that adheres to real-world physical laws’.


Chinese physical AI start-up proposes new paradigm that bypasses OpenAI, Meta road maps

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Shanghai-based Fysics AI says Fysiverse is a ‘new-generation physics-based world model that adheres to real-world physical laws’

Shanghai-based Fysics AI says Fysiverse is a ‘new-generation physics-based world model that adheres to real-world physical laws’