Data Makes the World Go 'Round

Chapter 15Building AI Models

In this chapter, we will examine the following:

• Building AI models
• Bias, fairness, and explainability
• Unstructured data

In the previous chapters, we examined some keys for AI success, including establishing a trusted data foundation and aligning AI strategy with business goals, and we touched upon data governance. We also talked about the roles required for AI. With some of this groundwork in place, we now turn to the heart of the matter: the AI itself. This chapter offers a high-level view of the processes behind building traditional AI models, such as machine learning. This includes training approaches and feature engineering as well as the algorithms that power common use cases such as churn prediction or customer retention.

We will explore how models are tested, what their outputs mean, and why concepts such as bias, fairness, and explainability are so critical for building trust. We’ll also look at how successful organizations overcome some of these issues. I think these principles are important so that you understand what data scientists do and what AI application developers need to know (at least at a high level).

This will set the stage for Chapter 16, which discusses how successful organizations operationalize their models, which is where the value of AI lies. This will also put into perspective what other teams (e.g., machine learning operations [MLOps]) do.

Building AI Models

As was previously mentioned, “artificial intelligence” ...