Monday, February 16, 2015

Worrisome message from the world of big data and recommendation systems.

Tapster - Similar Product Template: In this demo, we will show you how to build a Tinder-style web application recommending comics to users based on their likes/dislikes of episodes interactively.



I have just come across this tutorial 


http://docs.prediction.io/demo/tapster/

"In this demo, we will show you how to build a Tinder-style web application recommending comics to users based on their likes/dislikes of episodes interactively.”

I think what really got my attention was that at no point do we actually see any visualisation is going on. No attempt was made to try and demystify what the system was doing. This is as close as you can get to software as magic. What worries me is is this representative of the general process of big data processing? Okay with a recommenders system the worst it can do is make stupid suggestions but as a general process how do people spot that they're not making big mistakes?

Personally I think it is only the process of accurately visualising this kind of information that will allow people to see what errors than making 

No comments: