Object discovery is a core task in computer vision. While tremendous success has progressed in supervised object detection with vast annotated data, its unsupervised counterpart remains largely unexplored. With the growth of data volume, the expensive cost of annotations is the major limitation hindering further study. Therefore, discovering the objects without annotations has great significance. However, the task seems impractical on still-image or point cloud alone due to the lack of discriminative information. Previous studies overlook the crucial temporal information and constraints naturally behind multi-modality. In this paper, we propose 4D unsupervised object discovery, jointly discovering objects from 4D data——3D point clouds and 2D RGB images with temporal information. We present the first practical approach for this task by proposing a ClusterNet on 3D point clouds, which is joint iterative optimizing with a 2D localization network. Extensive experiments on the large-scale Waymo Open Dataset suggest that the localization network and ClusterNet achieve competitive performance on class-agnostic 2D object detection and 3D instance segmentation, bridging the gap between unsupervised methods and full supervision. Code will be released.