1.2 Describe How Various Data Types Are Used in Gen AI and the Business Implications. GENERATIVE-AI-LEADER Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   A startup wants to use a large public dataset of images to train a generative AI model. The dataset is hosted on a cloud storage service that charges per gigabyte for download, and the startup has a limited budget. Which data accessibility factor is most directly affecting their ability to use this data?

   • AAvailability
   • BCost
   • CFormat
   • DCompleteness
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A media company wants to build a gen AI model to generate video captions. They have a large archive of video files stored in a proprietary format that is expensive to access and process. The team is evaluating options to make the data more accessible for model training. Which approach best balances cost and accessibility?

   • APurchase additional licenses for the proprietary software to speed up access to the video files.
   • BConvert the videos to a widely supported open format like MP4, which is easier and cheaper to process with standard tools.
   • CUse only a small sample of videos for training to reduce the cost of accessing the proprietary format.
   • DOutsource the video processing to a third-party service that can handle the proprietary format, even if it increases costs.
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A multinational corporation wants to build a gen AI model to generate internal reports from both structured financial data (spreadsheets) and unstructured email communications. The company has strict data residency requirements: financial data must remain in the EU, while email data can be processed in the US. The team is evaluating cloud providers that offer gen AI services in both regions. Which consideration is most important when selecting the data storage and processing approach?

   • AStore all data in the EU region to ensure compliance, even if it increases costs for processing email data.
   • BUse a single global region with data replication to both EU and US, ensuring data is available in both locations.
   • CStore all data in the US region to simplify management, since the cloud provider offers strong security measures.
   • DStore and process financial data in an EU region and email data in a US region, ensuring compliance with data residency requirements.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A marketing agency wants to create a gen AI model that generates social media posts from a brand's historical content. The brand has a database of past posts (structured: date, platform, engagement metrics) and a collection of images and videos (unstructured). The agency has a limited budget and needs to decide how to combine these data types for training. Which approach is most effective?

   • AConvert the unstructured data into a structured format (e.g., tags) to simplify the training process.
   • BCombine both structured and unstructured data, using the structured data to inform the model about performance and the unstructured data to provide creative examples.
   • CUse only the unstructured data (images and videos) to train the model, as it contains the creative elements needed for generation.
   • DUse only the structured data (engagement metrics) to train the model, as it is easier to process and more reliable.
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A hospital is building a gen AI model to generate patient discharge summaries from electronic health records (EHRs). The EHRs contain structured data (lab results, medications) and unstructured data (physician notes). The hospital has a strict requirement for data privacy and cannot use cloud services that store data outside the country. The team is evaluating options to improve data quality while maintaining privacy. Which approach best balances data quality and privacy?

   • AUse a public cloud service in a different country to access more powerful gen AI tools, accepting the privacy risk.
   • BUse an on-premises solution to process the data, ensuring privacy, and implement data cleaning to improve completeness and consistency.
   • CAnonymize the data and use a public cloud service, as anonymization eliminates all privacy concerns.
   • DUse only the structured data (lab results, medications) to avoid privacy issues with unstructured notes.
   Reveal Solution Discussion Select an answer first