Fantasizing how AI works: Part 5.1

Objective: A humanized explanation of how a machine predicts and learns

In the previous blog (Part 4.1), we covered what a machine learns in mathematical terms as well as more humanized terms using words like intuition and gist.

We will now go through a multi-part fictional story to help explain how neural networks make predictions, which is closely tied to how they learn.

A Mystical Tour

A great tribe who call themselves the Oompa Kicchu live in the high mountains.

The Tribe of Oompa Kicchu (source)

Everyday, they deploy scouts in the surrounding Forest of Previously Unseen Beasts to protect the village from beastly threats.

Scouts (source)

The scouts have explicit instruction to send snapshots of the beasts to the High Tower of Propagated Wisdom, at the center of the village. There, many tribal folk work collaboratively to classify if the beast could be a threat to the village.

Tower of Propagated Wisdom (source)

In order to prepare future generations to work in the Tower of Wisdom, the tribal folk send their children to the Institute of Deep Learning. The specialists in the tower are all graduates of this institute.

Institute of Deep Learning (source)

Let’s take a tour of each of these attractions through this multi-part blog.

Allegory Unveiled

The Forest, the Tower of Wisdom, and Institute of Deep Learning are manifestations of neural network concepts.

• The beasts of the forest represent previously unseen data points that the neural network classifies as threats or not.
• The tower depicts neural networks, post-training.
• The institute depicts how neural networks are trained, before they occupy the ‘production ready’ tower.

Next stop:

Once Upon a Time: Part 5.2