According to the Financial Times, Elon Musk’s xAI has hired experts from Nvidia to build advanced world models; AI systems that can understand, simulate, and design physical environments.
The AI startup is taking on rivals like Meta and Google in developing next-generation AI that goes beyond text-based tools like ChatGPT or Grok. These models use data from videos and robots to learn how the real world behaves and could reshape how AI interacts with reality.
Two people familiar with xAI’s work reportedly said the company plans to use these world models in gaming, building fully interactive 3D environments that respond like real spaces. Also, the same systems could later power AI robots, training them to move and react naturally.
The company’s latest hires from Nvidia bring deep expertise in this area, making xAI a direct competitor in one of the most advanced AI fields yet.
xAI strengthens team with Nvidia researchers to expand gaming and robotics goals
xAI recruited Zeeshan Patel and Ethan He, two Nvidia researchers with experience developing simulation-based AI. Nvidia has long led the charge with its Omniverse platform, which allows developers to create detailed, physics-based simulations.
Some technology companies believe world models could eventually transform AI into something that functions across both digital and physical products, including humanoid robots.
Nvidia recently told the Financial Times that the potential global market for world models could reach the size of the entire world economy. That’s the scale of ambition now driving this technology. Elon said on X that xAI plans to launch an AI-generated game before the end of next year, confirming a goal he first mentioned in 2024.
The announcement sets a clear timeline for xAI’s gaming ambitions, tying together Musk’s push to merge simulation, creativity, and real-world interaction.
This week, xAI released a new image and video generation model with what it called “massive upgrades,” available for free. The company said the model creates sharper videos with more natural motion.
Unlike current systems like OpenAI’s Sora, which predicts frames from learned image data, world models are built to understand cause and effect – how gravity, objects, and space work in real time. This allows them to generate realistic movement and physics instead of just visuals.
xAI launches hiring spree and doubles down on real-world AI training
To expand its work, xAI is hiring technical staff for its “omni team,” which focuses on image, video, and audio generation. Job listings describe it as building AI experiences beyond text, including multimedia creation. The positions pay between $180,000 and $440,000 per year, depending on seniority.
There’s also a job posting for a “video games tutor,” who will train Grok to produce and design video games. That role pays $45 to $100 per hour, aimed at helping users explore AI-assisted game creation.
Elon joins Meta and Google in pushing toward world models, but the challenge remains steep. Training AI to replicate the real world requires massive datasets (footage, physics data, and 3D mapping) all of which are costly to gather and process. Even with access to Nvidia hardware and talent, xAI faces technical and financial hurdles to scale these systems effectively.
Not everyone in the gaming world is convinced. Michael Douse, publishing head at Larian Studios, the developer behind Baldur’s Gate 3, wrote on X that AI won’t solve the game industry’s “big problem,” which he said is about “leadership and vision.”
He added that the sector doesn’t need “more mathematically produced, psychologically trained gameplay loops,” but rather “expressions of worlds that people are actually engaged with or want to engage with.”
If you're reading this, you’re already ahead. Stay there with our newsletter.
