Multi-modal data refers to the training dataset itself, while multi-modal AI refers to the model trained on that data. You cannot have reliable multi-modal AI without high-quality, well-synchronized multi-modal data behind it.