Talk Abstract: In this talk, I will: 1) review recent work from our team on reconstructing and generating 3D worlds using radiance fields, 2) discuss how I think about the different varieties of radiance field techniques (NeRF, Instant NGP, 3DGS, etc) that are currently being explored, and how they relate to each other, and 3) scrutinize the relationship between explicitly 3D techniques like radiance fields and the increasingly powerful techniques for image and video generation that do not explicitly reason in 3D, which should cause us to re-evaluate the role of 3D in the modern AI era.

Speaker Biography: Jon Barron is a research scientist at Google DeepMind, where he works on computer vision and machine learning. He also recently moved to the Palo Alto area, and would appreciate advice on where the best coffee shop and sushi places are.