Data science can't be point and click

This article wasn't at all what I thought it was about.

To me, it's not automation that's the issue, it's how easy it is to get _wrong_ data out of this type of software. I think at a minimum you have to know the data well enough to ensure the numbers pass a "sniff test"; and hopefully, the user has a good knowledge of the schema they're pulling data from to know the pitfalls.

You don't have to be an expert, but knowing a little SQL to validate things can really help.

10 things statistics taught us about big data (talk by the same author, linked from the post): http://www.slideshare.net/jtleek/10-things-statistics-taught...

Yeah the heading is buzzfeed-like, but the content is great.

Yeah it's much better to have "experts" create our misleading stats.

The author seems to be mostly talking about misinterpreting results which is definitely a problem. I don't specifically see how "point and click" tools makes this problem worse (or better).

There's some finesse to using machine learning, but it's largely a blunt tool. There's no doubt that the research is going into automating predictive analysis.