News Brief
Elon Musk’s xAI Targets Gaming And Robotics With New ‘World Model’ AI Systems: Report
X (formerly Twitter) owner Elon Musk. (Image: Shutterstock)
Elon Musk’s artificial intelligence startup, xAI, is venturing into the next frontier of AI — the development of “world models", systems designed to navigate and design physical environments, Financial Times reported.
The move places Musk’s firm alongside leading players such as Google and Meta, which are also racing to merge digital intelligence with real-world physics.
Based in San Francisco, xAI has recruited Nvidia experts to develop these advanced systems that learn from robot data and video footage, aiming to simulate real-world understanding.
These world models could extend AI’s capabilities beyond large language models, which are trained on text and power tools like ChatGPT and xAI’s Grok.
xAI is reportedly developing world models for use in gaming to create interactive 3D environments. These models might also be used in AI systems for robots.
The company has reportedly hired AI researchers Zeeshan Patel and Ethan He from Nvidia, both of whom have experience in world model development.
Nvidia itself has led advancements in this field through its Omniverse platform, which builds and runs simulations.
Several technology groups have high expectations for world models, believing they could extend AI applications beyond software into physical products such as humanoid robots.
On Tuesday, xAI introduced its newest image and video generation model, describing it as having “massive upgrades” and making it freely available.
Existing video generation systems like OpenAI’s Sora produce video frames by predicting patterns that they have learned from training data.
World models would represent a major advancement by offering a causal understanding of physics and the interaction of objects within various environments in real time.
Musk’s company is following other leading AI laboratories, including Google and Meta, which are developing similar systems.
Despite their potential, world models remain a major technical challenge, as gathering enough data to simulate the real world and train these systems has been both difficult and expensive.
