I think this will hold true for the current interest around ML. We are already at a stage where we can not reliably 'debug' models. And feeding more data in

I think this will hold true for the current interest around ML. We are already at a stage where we can not reliably 'debug' models. And feeding more data in the pipeline seems to exhibit diminishing returns.