This week, at the All In conference, I got a glimpse of the secretive world of world models. Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs are the stars, but they’re not in it for the money. Their projects are still in the experimental phase, with no clear timeline for commercialization. Michael Rabbat, the VP of World Models at AMI, was surprisingly evasive when I asked what they were working on, saying they’ll reveal more when they’re ready.

Their secrecy even extends to their suppliers. Alex de Vigan, CEO of Physicl, is in the dark about AMI’s projects, wishing for more transparency. World models can be versatile, from self-driving cars to interactive video games, but AMI and World Labs are keeping their cards close to their chests. The reason? Delaying competition and staying under the radar.

It’s a bit like the Dark Forest scenario from Cixin Liu’s books, where the best strategy is to avoid drawing attention. As long as AMI and World Labs can keep a low profile, they can continue to innovate without the pressure of rivals. But this secrecy could also be holding back progress, as suppliers like Physicl could help more if they knew the direction of the projects.

For now, the world of world models remains a mystery, but one that could soon unravel as these companies near their breakthroughs. It’s an intriguing mix of ambition and caution, a testament to the complexities of AI innovation.