Inference 1: If the task is relatively fixed and there is a reliable solution, there is no need to call on a large model to take risks.
Inference 2: The tasks are not fixed, and adapting them one by one is too complex. The big model has a certain universality, and we need to make good use of this universality and entrust it to the big model.
Inference 3: The task is not fixed and can be adapted one by one, depending on the actual situation. If using a large model, it is necessary to consider the situation where the large model answers incorrectly and handle the errors accordingly.
Inference 4: The task is fixed and there is no reliable solution. If using a large model to attempt creative solutions, manual intervention is required.