The AI Product Playbook

CHAPTER 2How Machine Learning Models Learn: A Peek Under the Hood

You don't need to be a data scientist to manage AI products, but understanding the basics of how ML models learn is essential. This knowledge empowers you to collaborate effectively with your technical team, make informed product decisions, and set realistic expectations. We'll use a simple, relatable example: classifying fruits.

Imagine teaching a child to identify fruits. You show them Apples, Bananas, and Oranges, pointing out their differences. ML models learn similarly: they analyze examples and identify patterns.

“Features” are the specific characteristics the model uses to make predictions. Think of them as the descriptive attributes of your data. The example shown in Figure 2-1 uses the following features:

• Color: (Represented numerically: 1 = Red, 2 = Yellow, 3 = Orange)
• Shape: (Represented numerically: 1 = Round, 2 = Elongated)

Choosing the right features is key to building successful ML models. As a PM, you'll collaborate with your data science team to identify features that are:

• Relevant: They should have a strong connection to the outcome you're trying to predict.
• Available: You need to be able to collect and use this data.
• Understandable: Features should be relatively easy to interpret (even if the model itself is complex).

Figure 2-1: An illustration of fruits and features Color and Shape