Latent Space NYC

A Discussion

Emergent Reasoning in Video and World Models

Led by

Hokin Deng

Video generation models can now reason. Sora-2 achieves 60% on cognitive tasks once reserved for symbolic AI—chess, mazes, Sudoku, mental rotation, Raven's matrices.

We'll discuss Hokin's paper "Video Models Start to Solve Chess, Maze, Sudoku, Mental Rotation, and Raven's Matrices" and what it means for the field.

Curated group. Limited seats. By application only.