XD blog

blog page

big data, machine learning


2016-07-07 Machine Learning? What did you say?

Yesterday my sister asked : What is machine learning? And she quickly added please explain with terms I know.

Ok...

So I won't say this is just a bunch of optimization problems which apply on data to improve the efficiency of a company.

What should i say?

Should I apply the shaddock logic to my job: if I can't explain, I'm no use. Is it about quantifying people's intuition with data?

Well...

Machine Learning is everywhere but that won't help. In cars, in websites, in sales, but that won't help.

So I thought. Think about your job today, think about it in 20 years. A robot will probably do it. We always think a robot is a machine as if you could only automate physical jobs such as a mechanical arm. But the smaller robot is a program for computers. I press a button to update my blog: this is a robot. By extension, we could consider machine learning a kind of robot for statisticians. We don't try anymore to analyze the data, to model it. We just try to apply a black box which tells us if the data can be predicted. How? we care less.

Is that the end of the story?

The robot knows more and more and still can only solve problems we've seen. Not the last one. The new one. So the job is now to improve the robot so that I don't solve the same problem twice.


<-- -->

Xavier Dupré