12.30.2016

Google asked us to design illustrative gifs to help explain machine learning in the revamp of the “about” page on Google.com.

For the full Q&A head over to the newly designed google.com/about.

Neural net models learn from examples

Training a neural net to learn begins with the process of feeding it millions of labeled examples. In this scenario, the neural net will be shown photos of cats and dogs.

After the model is trained you can test it

After being trained with numerous labeled examples, we can send an unlabeled image through the neural network, and ask the machine to tell us what is in the image.

What does common sense look like to a machine?

In the field they call this “regularization”. The machine should still be able to understand that a dog with a cowboy hat is still a dog. If the example changes slightly, the machine should not totally change its mind. This is done by making the machine learning insensitive to small, unimportant changes.

If at first you don’t succeed, try another billion times

If there is a bumper sticker slogan for machine learning at Google, this is it.