At Luma AI, we're building multimodal general intelligence, starting with the world's best video models, and we've raised $200 million from investors like Andreessen Horowitz and Nvidia.

I left Apple to start Luma AI in 2022 after realizing that convincing a large organization to adopt new technology was unlikely.

AI video models are finally starting to work effectively, with new entrants like Hyo, Clling, and OpenAI's Sora making significant strides in the market.

The video diffusion model compresses hundreds of millions of internet videos into a simulation, enabling robots to interact with objects in a dream-like space.

We are not just a video model company; our goal is to achieve multimodal general intelligence, enabling users to create entire worlds.

The video shown was entirely generated by a custom model, illustrating the capabilities of fine-tuning general-purpose video generation models on specific domain data.

Models will get better at understanding and creating images, audio, and especially video, which will dramatically enhance how we interact with data.

During the first two years at Oculus, we experienced rapid changes and learning, managing to get our first development kit into customers' hands in just six months, which is not typical for hardware development.

OpenAI's latest model, o1, is gaining traction among developers, with 15% of the summer batch already using it despite its recent release.