Multimodal Representation and Embodied AI