Keep pulling the thread on Fei-Fei Li.
Fei-Fei Li asserts that 'world models' are the critical missing component in an AI landscape currently dominated by language models.
According to Martin, world models can take a 2D view of a scene and generate a complete 3D representation, including occluded parts not visible in the original view.
Fei-Fei Li is the founder and CEO of World Labs.
Martin credits Fei-Fei Li with singularly introducing the critical role of data into the field of AI.
World Labs defines a 'world model' as an AI that understands the 3D structure, shape, and compositionality of the physical world.
Fei-Fei Li founded World Labs because she believes developing world models requires a concentrated, industry-grade effort with significant compute, data, and talent.
Fei-Fei Li argues that the discovery of the 3D double helix structure of DNA is an example of a scientific breakthrough that required spatial reasoning and could not have been achieved using language alone.
Fei-Fei Li considers Neural Radiance Fields (NeRF) to be a significant revolution in the field of 3D computer vision.
The Neural Radiance Fields (NeRF) technique was developed by World Labs co-founder Ben Mildenhall and his colleagues at UC Berkeley approximately four years ago.
The pioneering work of World Labs co-founder Christoph Lassner was instrumental in popularizing Gaussian splatting as a method for representing volumetric 3D data.
World Labs co-founder Justin Johnson, a former student of Fei-Fei Li, is credited with doing foundational work in image generation.
Martin believes that World Labs' approach to creating world models requires a unique combination of expertise in both AI and computer graphics.