Skip to content
BytePatterns

AIF-C01 · Domain 1: Fundamentals of AI and ML · 20% of the exam

Task 1.2: Identify practical use cases for AI.

When AI earns its cost and when a fixed rule is the better answer, which technique fits which problem (regression, classification, clustering), and which managed AWS AI service already does the job.

Study it

  • When AI fits and when it does not: regression, classification, clustering

    Partly covered by: What Is Machine Learning

  • The managed AI services: Comprehend, Translate, Transcribe, Polly, Lex, Rekognition, Textract, Personalize

    Lesson coming

  • Traditional ML or a foundation model: explainability, regulation and operational limits

    Lesson coming

Sample questions

Try each one before opening the answer. Every option is explained, with the AWS documentation page that proves it.

Question 1 · choose 2

A company records its English-language support calls. It wants to turn each recording into text and then find out whether the customer's sentiment was positive, negative, neutral or mixed. Which AWS services should it combine? (Choose TWO.)

  1. AAmazon Polly
  2. BAmazon Transcribe
  3. CAmazon Textract
  4. DAmazon Comprehend
  5. EAmazon Translate
Show the answer and why

  • AAmazon Polly

    Incorrect

    Amazon Polly converts text into lifelike speech, the opposite direction of what the company needs.

  • BAmazon Transcribe

    Correct

    Amazon Transcribe is an automatic speech recognition service that converts audio to text.

  • CAmazon Textract

    Incorrect

    Amazon Textract detects and extracts text, forms and tables from documents, not from audio.

  • DAmazon Comprehend

    Correct

    Amazon Comprehend uses natural language processing on text and can determine its sentiment as positive, negative, neutral or mixed.

  • EAmazon Translate

    Incorrect

    Amazon Translate translates text between languages. The calls are already in one language, and translation does not score sentiment.

Speech to text is Transcribe; insight from text, including sentiment, is Comprehend. Chaining managed AI services like this needs no ML expertise.

AWS documentation

Task 1.2Report a problem with this question

Question 2 · choose 1

An online retailer has purchase histories for two million customers but no predefined customer categories. The marketing team wants to discover groups of customers who buy in similar ways. Which machine learning technique fits this goal?

  1. ALinear regression
  2. BBinary classification trained on labeled examples
  3. CClustering
  4. DMulticlass classification into existing segments
Show the answer and why

  • ALinear regression

    Incorrect

    Linear regression is a supervised technique that predicts a value on a continuous scale, such as a house price. It does not form groups.

  • BBinary classification trained on labeled examples

    Incorrect

    Classification predicts a known category and needs labeled training data. Here no categories exist yet; they have to be discovered.

  • CClustering

    Correct

    Clustering is an unsupervised technique that groups similar data inputs together. AWS lists grouping customers by purchasing behavior as a typical unsupervised learning use.

  • DMulticlass classification into existing segments

    Incorrect

    Multiclass classification assigns each input to one of several known labels. The retailer has no segment labels to train on.

No labels and a goal of discovering groups points to unsupervised learning, and specifically to clustering.

AWS documentation

Lesson: What Is Machine LearningTask 1.2Report a problem with this question

Question 3 · choose 1

A video streaming service wants to show each viewer a "Top picks for you" row based on that viewer's watch history. The team has no machine learning experience and wants a managed service built for this. Which AWS service should it use?

  1. AAmazon Personalize
  2. BAmazon Comprehend
  3. CAmazon Lex
  4. DAmazon Rekognition
Show the answer and why

  • AAmazon Personalize

    Correct

    Amazon Personalize is a fully managed ML service that uses your data to generate item recommendations for your users, with video recommendations such as "Top picks for you" as a documented use case.

  • BAmazon Comprehend

    Incorrect

    Amazon Comprehend extracts insights such as entities, key phrases and sentiment from text. It does not generate recommendations.

  • CAmazon Lex

    Incorrect

    Amazon Lex builds conversational interfaces using voice and text, such as chatbots.

  • DAmazon Rekognition

    Incorrect

    Amazon Rekognition analyzes images and video, for example to detect objects or compare faces.

Recommendations from interaction history are the job Amazon Personalize was built for; the other three services analyze language or images.

AWS documentation

Task 1.2Report a problem with this question

Question 4 · choose 1

An insurer receives thousands of scanned claim forms every day, some filled in by hand. It wants to pull values such as the policy number and the rows of an itemized costs table out of each form and into its claims database. Which AWS service is designed for this?

  1. AAmazon Comprehend
  2. BAmazon Rekognition
  3. CAmazon Transcribe
  4. DAmazon Textract
Show the answer and why

  • AAmazon Comprehend

    Incorrect

    Amazon Comprehend extracts insights such as entities, key phrases and sentiment from text. It is not built to pull form fields and table rows out of scanned documents.

  • BAmazon Rekognition

    Incorrect

    Amazon Rekognition is an image and video analysis service for objects, faces and unsafe content. Extracting forms and tables from documents is what Amazon Textract is documented for.

  • CAmazon Transcribe

    Incorrect

    Amazon Transcribe converts audio to text. The claims arrive as scanned documents, not recordings.

  • DAmazon Textract

    Correct

    Amazon Textract detects typed and handwritten text in documents and extracts text, forms and tables with its Document Analysis API.

Forms, tables and handwriting in scanned documents point to Amazon Textract.

AWS documentation

Task 1.2Report a problem with this question

Question 5 · choose 1

A clinic wants customers to book and change appointments by talking or typing to a chatbot on its website and phone line. The team has no deep learning expertise. Which AWS service is built for this?

  1. AAmazon Lex
  2. BAmazon Polly
  3. CAmazon Textract
  4. DAmazon Personalize
Show the answer and why

  • AAmazon Lex

    Correct

    Amazon Lex builds conversational interfaces using voice and text, with natural language understanding and speech recognition, and needs no deep learning expertise.

  • BAmazon Polly

    Incorrect

    Amazon Polly turns text into lifelike speech. It can give a bot a voice but does not understand or manage a conversation.

  • CAmazon Textract

    Incorrect

    Amazon Textract extracts text, forms and tables from documents.

  • DAmazon Personalize

    Incorrect

    Amazon Personalize generates item recommendations from user data.

Conversational interfaces by voice and text are Amazon Lex; Polly only gives an application a voice.

AWS documentation

Task 1.2Report a problem with this question

Question 6 · choose 1

An online store wants to show its product descriptions to shoppers in ten languages and keep the translations in step as descriptions change. Which AWS service fits?

  1. AAmazon Polly
  2. BAmazon Translate
  3. CAmazon Comprehend
  4. DAmazon Transcribe
Show the answer and why

  • AAmazon Polly

    Incorrect

    Amazon Polly converts text into speech in many languages; it does not translate text from one language into another.

  • BAmazon Translate

    Correct

    Amazon Translate is a text translation service for translating documents and building applications that work in multiple languages.

  • CAmazon Comprehend

    Incorrect

    Amazon Comprehend can detect a document's language and extract insights, but it does not translate the text.

  • DAmazon Transcribe

    Incorrect

    Amazon Transcribe converts speech in audio to text.

Text in one language to text in another is Amazon Translate.

AWS documentation

Task 1.2Report a problem with this question

Question 7 · choose 1

A marketplace lets sellers upload product photos. It wants to flag images with inappropriate or offensive content automatically so that human moderators review only the flagged ones. Which AWS service fits?

  1. AAmazon Comprehend
  2. BAmazon Textract
  3. CAmazon Personalize
  4. DAmazon Rekognition
Show the answer and why

  • AAmazon Comprehend

    Incorrect

    Amazon Comprehend extracts insights from text. It does not moderate image content.

  • BAmazon Textract

    Incorrect

    Amazon Textract extracts text, forms and tables from documents; it does not judge whether an image is offensive.

  • CAmazon Personalize

    Incorrect

    Amazon Personalize recommends items to users. It does not inspect image content.

  • DAmazon Rekognition

    Correct

    Amazon Rekognition moderation APIs detect inappropriate, unwanted or offensive content in images and videos, so human moderators can focus on a smaller set of content.

Image and video moderation is a computer vision task, and Amazon Rekognition is the managed AWS service for it.

AWS documentation

Task 1.2Report a problem with this question

Question 8 · choose 1

Which use case is better served by a traditional machine learning model trained on the company's own data than by a general-purpose foundation model?

  1. ADrafting personalized marketing emails in several different tones of voice
  2. BAnswering open-ended employee questions in a chat
  3. CPredicting churn from account data with explainable results
  4. DSummarizing long reports into short briefings
Show the answer and why

  • ADrafting personalized marketing emails in several different tones of voice

    Incorrect

    Generating varied text is a general task foundation models perform well out of the box.

  • BAnswering open-ended employee questions in a chat

    Incorrect

    Conversing in natural language is one of the general tasks foundation models are trained to perform.

  • CPredicting churn from account data with explainable results

    Correct

    Churn prediction on structured data is a classic predictive ML task, and when each decision must be explained, for example to regulators, a model whose behavior can be explained is required.

  • DSummarizing long reports into short briefings

    Incorrect

    Summarization is a general language task foundation models handle well without training a task-specific model.

Foundation models shine at general language and content tasks. A narrow prediction on structured data, with explainability or regulatory demands, often fits a traditional ML model better.

AWS documentation

Task 1.2Report a problem with this question

Question 9 · choose 2

A company is listing projects for its first machine learning initiative. Which problems are good candidates for ML? (Choose TWO.)

  1. AFlagging likely fraudulent insurance claims among millions of claims
  2. BAdding a fixed, published shipping fee to every order total
  3. CForecasting demand for each product in each store next month
  4. DSorting a customer list alphabetically by last name
  5. EConverting order amounts from cents to dollars
Show the answer and why

  • AFlagging likely fraudulent insurance claims among millions of claims

    Correct

    Insurance fraud detection is one of the ML applications AWS lists; ML finds patterns in large amounts of historical data.

  • BAdding a fixed, published shipping fee to every order total

    Incorrect

    A known, exact rule needs no learning; ML is for tasks performed without explicit instructions, relying on patterns and inference.

  • CForecasting demand for each product in each store next month

    Correct

    Retail demand forecasting is another ML application AWS lists.

  • DSorting a customer list alphabetically by last name

    Incorrect

    Sorting follows an explicit, exact procedure. There is no pattern to learn from data.

  • EConverting order amounts from cents to dollars

    Incorrect

    A fixed arithmetic conversion is fully specified by a rule; nothing is gained by learning it from examples.

ML earns its cost where rules are hard to write and patterns live in data. Where an exact, known rule exists, plain code is cheaper and exact.

AWS documentation

Lesson: What Is Machine LearningTask 1.2Report a problem with this question

Practise domain 1 →Practise all domains →