How do memory and storage requirements differ between training and inference in AI systems?

The correct answer highlights the essential differences between training and inference in AI systems, particularly in terms of memory and storage requirements. During training, AI models are developed using large datasets, which require significant amounts of memory and storage. This is because training involves not only processing vast amounts of input data but also maintaining additional states such as gradients, parameters, and various internal weights that evolve throughout training iterations. As a result, the need to store this information while adjusting the model makes the memory and storage requirements for training considerably high.

On the other hand, inference requires less memory and storage compared to training. During inference, the model has already been trained, and it primarily involves applying the learned parameters to new input data. This process does not require the same level of resource intensity since it does not involve updating model weights or maintaining state as during training.

This understanding clarifies why option C is the most accurate representation of AI systems' operational needs during training versus inference.