Difference between data centric and model centric?

1 question which just comes to my mind is that, what is the difference between data centric AI and normal functioning. Because I think we were already going for data centric approach, that’s why we had data scientists, right???


The point here is to build smarter models rather than collecting large amounts of data (but crude data) fed into the model. If the data fed into the model encompasses a variety of possible scenarios i.e. is chosen so (be it even in small quantity but represantative of the phenomena) and the model is well-built, it can learn from it well. On the other hand even if you have a lot of data but not well chosen and the model is not able to exctract the most important features, then the model learning will be distorted (plus it takes a lot of time to process learning huge amounts of data). Ultimately its quantity vs quality, a better model with well choosen training data may perform as good or even better than a model with huge dataset.

Hope it helps.