The World Model Mirage: Venture Capital's Newest Black Box

In the current AI gold rush, a bold new term emerges, venture capital floods in, and technical specifics often vanish into a cloud of vague promises. We are seeing this play out in real-time with the rise of 'world models,' a trend detailed in recent reporting by TechCrunch.

On the surface, the pitch is seductive. World models are designed to automate 'spatial intelligence,' with theoretical applications ranging from humanoid robots and self-driving systems to interactive video and CGI effects. This versatility suggests a total addressable market that encompasses nearly every physical industry.

But behind the hype lies a lack of transparency. At the All In conference, TechCrunch’s Russell Brandom attempted to pin down the commercial trajectory of these companies, only to be met with silence. Michael Rabbat, VP of World Models and co-founder of Yann LeCun’s AMI Labs, was notably cagey, telling Brandom, “We’ll talk about it when we’re ready to talk about it.” Rabbat later clarified via email that AMI Labs is in a 'research and building phase' and is not disclosing product plans.

This caginess is systemic. Alex de Vigan, CEO of data supplier Physicl, told Brandom that while his data is being used, he has no idea what the companies are actually building. De Vigan noted that Physicl could provide more useful data if clients disclosed their objectives.

Even when products exist, they remain vague. World Labs, founded by Fei-Fei Li, has a product called Marble that creates explorable environments for games and CGI. However, Brandom observes that the platform seems designed to demonstrate general capabilities rather than a specific business model. Similarly, AMI Labs has experimented with everything from robotics and manufacturing to AI software for doctors via a partnership with Nabia—a scattergun approach that suggests a lack of profitability focus.

TechCrunch suggests this is a 'dark forest' scenario: by staying vague, labs avoid immediate competition from rivals like OpenAI and Anthropic while continuing to fundraise. Until the leaders of AMI Labs and World Labs provide a concrete product that solves a specific problem, 'world models' remain a theoretical curiosity. Versatility without a product is just a fancy word for having no plan.