The dark forest of world models: why the labs stay quiet

TechCrunch's Russell Brandom moderated a panel on world models at the All In conference and used it to probe one of the most secretive corners of the AI industry. The big names here are Yann LeCun's AMI Labs and Fei-Fei Li's World Labs. Both have accumulated plenty of buzz and funding — and both rank near the bottom on any scale measuring an actual effort to make money.

At heart, world models are about automating spatial intelligence, which opens many lucrative directions: robotics, interactive video, more sophisticated self-driving systems. But the moment the conversation turns to where the technology will actually be commercialized, things go foggy.

"We'll talk when we're ready"

The closest thing to an authority was Michael Rabbat, AMI Labs co-founder and VP of World Models. Pressed on what the company is building, he was cagey: "We'll talk about it when we're ready to talk about it." Over email he clarified that the firm is still in a research-and-building phase and won't discuss products or timelines publicly. AMI is less than a year old, so the silence is fair enough.

But the reticence runs through the whole field. World Labs' Marble looks like the most developed product around, with demos spanning media creation, explorable environments for games, and CGI effects. There are robotics use cases too, yet the platform seems built to showcase capabilities rather than sell a specific solution.

Secrets even from suppliers

The secrecy extends to vendors. Alex de Vigan, CEO of Physicl — a data supplier for the emerging world-model business — admitted he doesn't know what his own clients are making. "I wish they would tell us more. We could build more useful data if we knew what they were working on," he said.

Part of the mystery lies in how versatile the idea is. The simplest version is a navigable map of the world, like the models powering self-driving cars. But the same approach can help a humanoid robot carry boxes or turn a few minutes of footage into an explorable space. AMI has already dipped into manufacturing, biomedicine, robotics, and even software for doctors.

Dark forest

As long as fundraising stays easy, there is little pressure to pick a single focus — and good reason not to. If AMI announced tomorrow that it had built a humanoid robot or a next-generation rendering system, rivals would swarm: other world-model labs, the neolabs, and even OpenAI and Anthropic. The flip side of easy money is that your competitors can raise it too.

Brandom compares the situation to the dark-forest scenario from Cixin Liu's novels: if you don't know who else is in the woods, it's best not to draw attention to yourself.

Source: TechCrunch