That is already the case for most neural networks and LLMs.