Book Image

Microsoft Azure AI Fundamentals AI-900 Exam Guide

By : Aaron Guilmette, Steve Miles
Book Image

Microsoft Azure AI Fundamentals AI-900 Exam Guide

By: Aaron Guilmette, Steve Miles

Overview of this book

The AI-900 exam helps you take your first step into an AI-shaped future. Regardless of your technical background, this book will help you test your understanding of the key AI-related topics and tools used to develop AI solutions in Azure cloud. This exam guide focuses on AI workloads, including natural language processing (NLP) and large language models (LLMs). You’ll explore Microsoft’s responsible AI principles like safety and accountability. Then, you’ll cover the basics of machine learning (ML), including classification and deep learning, and learn how to use training and validation datasets with Azure ML. Using Azure AI Vision, face detection, and Video Indexer services, you’ll get up to speed with computer vision-related topics like image classification, object detection, and facial detection. Later chapters cover NLP features such as key phrase extraction, sentiment analysis, and speech processing using Azure AI Language, speech, and translator services. The book also guides you through identifying GenAI models and leveraging Azure OpenAI Service for content generation. At the end of each chapter, you’ll find chapter review questions with answers, provided as an online resource. By the end of this exam guide, you’ll be able to work with AI solutions in Azure and pass the AI-900 exam using the online exam prep resources.
Table of Contents (20 chapters)
Free Chapter
1
Part 1: Identify Features of Common AI Workloads
4
Part 2: Describe the Fundamental Principles of Machine Learning on Azure
8
Part 3: Describe Features of Computer Vision Workloads on Azure
11
Part 4: Describe Features of Natural Language Processing (NLP) Workloads on Azure
14
Part 5: Describe Features of Generative AI Workloads on Azure

Identify Responsible AI considerations for generative AI

Microsoft has created a framework for responsible AI and generative AI solutions comprised of four stages:

  • Identify potential harms that could be related to your planned solution
  • Measure the presence of those identified harms in the solution’s output
  • Mitigate the harms at multiple levels to minimize their expression and impact
  • Operate the solution responsibly

Let’s look at each of those four areas.

Identify

The first stage in implementing a responsible generative AI solution is to identify potential harms that may result from your solution.

Identifying potential harms or risks

You must identify possible risks associated with your generative AI project, which vary based on the services, models, and data you employ. Here are some common risks:

  • Generating offensive or biased content
  • Spreading misinformation
  • Promoting harmful behavior

To understand the limitations...