1.2 Describe How Various Data Types Are Used in Gen AI and the Business Implications. GENERATIVE-AI-LEADER Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   A team is training a supervised machine learning model to classify emails as 'spam' or 'not spam'. What type of data do they need for this task?

   • ASemi-structured data, because emails have headers and body
   • BUnlabeled data, because the model will learn to label the emails itself
   • CStructured data, because the model needs a table of features
   • DLabeled data, because each email must have a known 'spam' or 'not spam' label for training
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A healthcare organization wants to use gen AI to extract insights from clinical notes, lab reports, and physician dictations. The data is stored in various formats: some in a structured database, some in PDFs, and some as audio recordings. The team is concerned about the cost and time required to prepare this data for model training. Which approach best addresses the variety of data types while managing cost and effort?

   • AUse separate models for each data type and combine their outputs, leveraging specialized models that handle structured, unstructured, and audio data effectively.
   • BFocus only on the structured data because it is easier to process and yields more reliable results, ignoring the unstructured data for now.
   • CConvert all data to a single structured format before training, ensuring consistency and reducing preprocessing effort.
   • DHire a large team of annotators to manually label all data, ensuring high quality and consistency across all formats.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   A company wants to analyze customer feedback from social media posts to train a generative AI model. Which type of data are they primarily working with, and why?

   • AUnstructured data, because social media posts are free-form text without a predefined schema
   • BStructured data, because social media posts are stored in a database
   • CSemi-structured data, because posts have metadata like timestamps
   • DLabeled data, because the posts are tagged with user names
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A financial services firm is training a gen AI model to summarize regulatory documents. They have collected a large corpus of documents from various sources, but some documents are missing sections, others contain contradictory information, and many are not relevant to the regulatory domain. The team needs to improve the quality of the training data. Which action should they prioritize?

   • AConvert all documents to a structured format such as JSON to improve consistency and make them easier to process.
   • BFocus on cleaning the data by removing irrelevant documents, filling in missing sections, and resolving contradictions.
   • CIncrease the volume of data by adding more documents from the internet to ensure the model has enough examples.
   • DUse the data as-is because gen AI models are robust to noise and will learn to ignore irrelevant information.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A startup is building a gen AI chatbot for customer support. They have a large dataset of customer service conversations, but only a small portion is labeled with the correct response type. The team wants to improve the model's performance without incurring high labeling costs. Which strategy should they adopt?

   • AUse only the labeled data for training, as unlabeled data cannot be used for generative models.
   • BManually label a larger portion of the data to ensure the model has enough supervised examples.
   • CUse the labeled data to fine-tune a pre-trained model and use the unlabeled data for self-supervised learning to improve language understanding.
   • DDiscard the unlabeled data to avoid introducing noise into the model.
   Reveal Solution Discussion Select an answer first