Why Are World Model Companies So Tight-Lipped About Trade Secrets?

Overview

At the recent All In conference, discussions surrounding world models drew widespread attention from the industry. Although leading players such as AMI Labs, led by Yann LeCun, and World Labs, founded by Fei-Fei Li, have accumulated massive funding and public buzz, their commercial monetization remains relatively low. This phenomenon stems from the industry's high degree of caution regarding technological pathways and trade secrets.

The Fog of Commercialization

The core of world models lies in automated spatial intelligence, with application scenarios holding immense potential, spanning robotics, interactive video, and advanced autonomous driving systems. However, when asked about specific commercialization timelines, responses within the industry are often vague.

"We will announce when we are ready."

Michael Rabbat, Vice President of World Models at AMI Labs, responded cautiously to this during a panel discussion. He further clarified that the company is still in the research and development phase and has no plans to release public product plans or timelines. While this confidentiality strategy is understandable—since AMI was founded less than a year ago—it reflects the entire sector's lockdown on details.

Currently, World Labs' Marble platform is arguably the most mature product in the field, with demos covering media creation, game environment construction, and CGI effects. Nevertheless, these demonstrations focus more on proving technical capabilities rather than providing a clear path to commercial implementation.

The Information Dilemma in the Supply Chain

This culture of secrecy even extends to upstream suppliers. Alex de Vigan, CEO of Physicl and a data supplier for world model business, bluntly expressed his confusion. He acknowledged that Physicl's data has been proven effective but stated that without knowing the customers' specific building goals, they cannot optimize data production accordingly.

"We hope to learn more. If we knew what they were working on, we could build more useful data," said de Vigan.

Strategic Choices in the "Dark Forest"

The universality of world models is the primary reason for the silence. The same modeling logic can be applied to autonomous driving, humanoid robot handling, or exploratory environments for video generation. AMI has already experimented in multiple fields including manufacturing, biomedicine, robotics, and medical AI, but it is obviously impossible to expand across all of them.

In an environment where capital is abundant, companies lack the pressure to focus on a single track. Conversely, disclosing technical breakthroughs (such as humanoid robots or next-generation rendering systems) too early might attract competitors, emerging labs, and even giants like OpenAI and Anthropic.

This is just like the Dark Forest law depicted in Liu Cixin's works: when the locations of other hunters are unknown, exposing oneself is the most dangerous behavior. For world model companies, delaying the onset of competition is the best strategy, making tight-lipped silence a rational instinct for business survival.