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.