Learning Objectives
By the end of this chapter, you will be able to:
- Distinguish AI, machine learning, generative AI, foundation models, multimodal foundation models, diffusion models, and large language models, and explain when prompt engineering versus prompt tuning is the right intervention.
- Classify data as structured or unstructured and labeled or unlabeled, walk through the five stages of the ML lifecycle, and map each stage to the Google Cloud product that supports it.
- Describe the five layers of the gen AI ecosystem and select a foundation model against modality, context window, security, availability and reliability, cost, performance, fine-tuning, and customization criteria.
- Explain how Google's AI-first approach changes product design, and identify when content creation, summarization, discovery, automation, personalization, or recommendation is the right gen AI use case for a business problem.
Executive Summary
- Generative AI is a specific subset of machine learning that produces new content (text, image, audio, video, code) rather than only classifying or predicting from existing data. Foundation models are the pretrained engines under that capability.
- The ML lifecycle is the operating model leaders should plan against. Each of its five stages has a Google Cloud product anchor, and procurement, governance, and FinOps conversations all key off that mapping.
- The gen AI ecosystem stacks into five layers (infrastructure, models, platforms, agents, gen AI-powered applications). Knowing which layer a vendor sits on is the cleanest way to decide build versus buy.
- Google's AI-first approach is not a marketing slogan; it is a research, product, and infrastructure choice that shows up as Gemini, Vertex AI, custom AI accelerators (TPUs), and a published responsible-AI framework. Business value lands in six recurring use-case shapes that the exam tests directly.
Assumptions
- The reader has general IT literacy (knows what a database is, what an API does, what cloud computing is) but is not assumed to be a data scientist or ML engineer.
- Google Cloud product names follow the official naming the provider publishes; legacy names (for example, Cloud Functions, Data Studio) are noted only where the exam still references them.
- Numeric model limits change with each model release. Limit callouts in this chapter quote the value published in the official documentation as a stable parameter, and call out variability where the figure shifts release to release.
- Region availability of specific services is not exhaustively listed; assume a flagship Google Cloud region (for example,
us-central1) unless a section says otherwise.