No , No and No. The only thing that will happen is that the training data for those model will diminish. Note that you would not run out of data, but you will r

No , No and No. The only thing that will happen is that the training data for those model will diminish. Note that you would not run out of data, but you will run out of training data. As this happen all models will converge. Another path: you will get more and more specialized models that would be fed by specific domain data, and will always bit the more "compute rich" models.