Recent developments in large language models have propelled advancements in artificial intelligence, yet integrating multiple modalities remains challenging. Multimodal frameworks, which combine different types of data inputs, face significant obstacles such as motion quantization and cross-modal interference. These issues complicate the effective processing and interpretation of diverse data streams. As the number of modalities incorporated into these systems increases, the complexity of managing interactions and ensuring coherent outputs grows substantially. This highlights the intricate balance required to advance multimodal AI without compromising performance. The discussed article emphasizes these challenges within the context of human motion as a secondary modality, underscoring the ongoing efforts to refine such frameworks. Overall, the integration of additional modalities demands careful consideration to address the inherent technical difficulties.