1.2 Describe How Various Data Types Are Used in Gen AI and the Business Implications. GENERATIVE-AI-LEADER Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A research institute is building a gen AI model to generate scientific paper abstracts. They have collected papers from various journals, but some papers are missing the abstract, others have incomplete references, and some are not in the field of interest. The team needs to ensure the model produces high-quality abstracts. What should they do?

   • AFilter the dataset to include only papers that are complete, relevant, and have abstracts, to ensure high data quality.
   • BUse all papers as-is, including those without abstracts, to increase the volume of training data.
   • CConvert all papers to a standard format like PDF to ensure consistency in processing.
   • DManually write abstracts for the papers that are missing them to ensure completeness.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   A research team needs to access a proprietary dataset that is stored in a legacy system only accessible during business hours. The team needs to run training jobs overnight. Which data accessibility factor is most directly hindering their work?

   • AAvailability
   • BFormat
   • CRelevance
   • DCost
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A retail company wants to build a gen AI model to generate product descriptions from images. They have millions of product images, but only a few thousand have descriptive text. The team is deciding how to use this data for training. Which strategy is most effective for leveraging the available data?

   • ADiscard the unlabeled images because they lack the necessary annotations for training a generative model.
   • BUse only the labeled images for training, as they provide the ground truth needed for supervised learning.
   • CUse the labeled images for supervised fine-tuning and the unlabeled images for self-supervised pretraining to improve the model's understanding of product features.
   • DManually label all images to ensure the model has sufficient supervised data, even though it will take significant time and cost.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A non-profit organization wants to use gen AI to analyze survey responses from different regions. The surveys are collected in multiple formats: some are paper-based (scanned images), some are online forms (structured data), and some are audio recordings of interviews. The organization has a limited budget and needs to make the data accessible for model training. Which approach is most cost-effective?

   • AHire a third-party service to process all the data, even though it will increase costs significantly.
   • BDigitize the paper surveys and transcribe the audio recordings, then combine all data into a unified text format for analysis.
   • CUse only the online form data because it is already structured and requires no additional processing.
   • DUse the paper surveys as-is, since they are the most reliable source of data.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   Which of the following is an example of unstructured data?

   • AA JSON file with product information
   • BA collection of customer support emails
   • CA spreadsheet of customer names and purchase amounts
   • DA database table of employee records
   Reveal Solution Discussion Select an answer first