Cloud Technology and Services · Topic 8 of 11
AI and Machine Learning Services
Match AWS AI and machine learning services to tasks for CLF-C02, including SageMaker, Bedrock, Rekognition, Comprehend, Lex, Polly and Textract.
- 4 min
- SageMaker builds, trains and deploys custom ML models.
- Bedrock offers foundation models for generative AI through a managed service.
- Rekognition analyses images and video. Textract extracts text and data from documents.
What it is
Artificial intelligence (AI) is software that performs tasks that normally need human judgement, such as understanding speech or recognising images. Machine learning (ML) is a way to build such software by training a model on data instead of writing fixed rules.
AWS offers ML in layers.
At the top are ready made AI services you call like any other service. In the middle is a platform for building your own models. Below are the raw compute resources.
The exam does not ask you to build models. It asks you to pick the service that fits a described task.
How it works
Building your own models
Amazon SageMaker is a managed platform to build, train and deploy ML models. Choose it when you have your own data and need a custom model.
It saves effort compared with setting up servers and tools yourself, but you still need ML skills. If a ready made AI service already does the task, that is the simpler choice.
Generative AI
Generative AI creates new content such as text, images or code. Amazon Bedrock gives access to foundation models, which are large pre trained models, from AWS and other providers through one managed service. You can use them in your applications without training your own.
Amazon Q is a generative AI assistant. Amazon Q Business answers questions about a company's own data, and Amazon Q Developer helps developers with code and AWS tasks.
Ready made AI services
| Service | What it does |
|---|---|
| Amazon Rekognition | Analyses images and videos, such as finding objects, faces or text |
| Amazon Comprehend | Natural language processing: finds sentiment, key phrases and entities in text |
| Amazon Lex | Builds chatbots and voice bots with conversation |
| Amazon Polly | Turns text into spoken audio (text to speech) |
| Amazon Transcribe | Turns speech into text (speech to text) |
| Amazon Translate | Translates text between languages |
| Amazon Textract | Extracts text, forms and tables from scanned documents |
| Amazon Kendra | Intelligent search across a company's documents |
| Amazon Personalize | Provides recommendations for users, such as products to show |
Many of these names are close to the task they perform. Polly speaks, Transcribe writes down speech, Translate converts languages, and Textract pulls text out of documents.
Choosing between services
| Need | Service |
|---|---|
| Detect objects or faces in photos | Rekognition |
| Read text out of scanned invoices and forms | Textract |
| Understand sentiment in customer reviews | Comprehend |
| Build a chatbot | Lex |
| Create voice from text | Polly |
| Create text from voice | Transcribe |
| Translate a website | Translate |
| Recommend products | Personalize |
| Search internal documents with natural language | Kendra |
| Build a custom model with your data | SageMaker |
| Use a foundation model in an app | Bedrock |
Why use managed AI
Managed services let teams add AI features without ML expertise, and without buying special hardware. You pay for what you use, and AWS runs the infrastructure.
Model output can be wrong or biased, so people should review important results before acting on them.
Shared responsibility
AWS secures the AI infrastructure. You are responsible for your data, who can call the service, and for checking that outputs are suitable for your use.
What to remember for the exam
- SageMaker builds, trains and deploys custom ML models.
- Bedrock offers foundation models for generative AI through a managed service.
- Rekognition analyses images and video. Textract extracts text and data from documents.
- Comprehend analyses text meaning, such as sentiment and entities.
- Lex makes chatbots. Polly is text to speech. Transcribe is speech to text. Translate translates.
- Kendra is intelligent document search. Personalize makes recommendations.
- Amazon Q is a generative AI assistant for business data or for developers.
- Prebuilt AI services need no ML expertise.
Turn each service name into its job. Polly speaks, Transcribe writes down what was said, and Translate changes the language.
Common mistakes
Choosing SageMaker for a simple task. A prebuilt service such as Rekognition is faster if it already does the job.
Mixing up Polly and Transcribe. Polly turns text into speech. Transcribe turns speech into text.
Confusing Textract with Rekognition. Textract is built for documents and forms, while Rekognition is built for images and video content.
Assuming AI output is always correct. Models can make mistakes, so important results need review.
Bedrock gives you access to existing foundation models. It is not the same as SageMaker, which you use to build and train your own models.
Check yourself
Which service converts speech to text?
Amazon Transcribe.
Which service extracts text and tables from scanned documents?
Amazon Textract.
Which service provides foundation models for generative AI applications?
Amazon Bedrock.
Topic 8 of 11 in Cloud Technology and Services
Test yourself on Cloud Technology and Services
The full AWS CLF-C02 question bank has 832 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by AWS. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.