The "blackbox" approach for machine learning is the same concept Simon Chan used to build Prediction.io
The bet there is "would developers be fine with using a blackbox?". Or would they prefer to have full control on the algo inside the box?
Nice reading tho.
I believe "common" developers are totally fine with it, they always need some sort of abstraction layer. On the contrary, the (quickly growing) data scientists community is not: they need full control for their specific use cases and will never blindly use a black box API.
