Data Makes the World Go 'Round

Chapter 17Generative AI

In this chapter, we will examine the following:

• Understanding generative AI
• Foundation models
• AI consumers and builders
• Putting generative AI models into production with company data
• Hallucinations, bias, and guardrails
• Measuring success with generative AI

Up until now, I’ve been discussing data and AI from the perspective of the way data and analytics professionals see it. However, with the advent of generative AI (GenAI) a few years ago, the market has shifted into uncharted territory—fast-moving and disruptive, with organizations lacking clear guardrails compared to the more structured world of traditional data and AI.

There are different ways to look at this. On the one hand, this is a time of great innovation. Companies are creating new foundation models, building open systems, releasing new AI applications, and experimenting with novel use cases. That is exciting. In the Expert Advice from Daniel Ziv later in this chapter, we’ll hear that perspective directly. Some people say that with GenAI, the data doesn’t need to be perfect, just good enough. The advice is to try out the tools, see what happens, and fail fast. There is merit to that mindset. It encourages experimentation and lowers the barrier to entry. There is a sense of excitement around that, and much of it is well deserved.

On the other hand, when I speak to data and analytics professionals, I hear enthusiasm but caution. Their experience tells them that ignoring data quality, governance, ...