If an AI comes in the form of a trained model that is stateful but iterative (each generation of the model remains the same but can influence the future generation of its own model by adding training data, and humans still review the data sets and initiate the training, and the AI exists only processes in a discrete request/response transactional model, such that if nobody asks it a question, it will remain indefinitely idle, could it ever meet our definitions of being "alive"?