Generative AI and Large Language Models (LLMs) AIE Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   What is a major ethical concern associated with generative AI?

   • AIt eliminates all privacy risks
   • BIt can be used to create convincing fake news or deepfakes
   • CIt always reduces the need for human oversight
   • DIt guarantees unbiased outputs
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A design team wants to generate unique marketing images for a campaign. They have a generative AI model that can create images from text descriptions. They want to ensure the images are diverse and do not contain offensive content. Which approach is most effective?

   • AGenerate images with no prompts and rely on the model's default behavior.
   • BUse carefully crafted prompts that specify the desired style and explicitly exclude offensive content, and review the outputs before use.
   • CUse a text-only LLM to describe images instead of generating them.
   • DUse a discriminative model to select images from a pre-existing library.
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A social media company is deploying an LLM chatbot. They are concerned about the chatbot generating harmful misinformation. Which combination of measures is most effective?

   • AOnly use a larger LLM, as it will be more accurate and safe.
   • BDisable the chatbot's ability to learn from user interactions.
   • CIncrease the temperature to make the chatbot more creative and less likely to repeat misinformation.
   • DUse prompt engineering to restrict topics and implement a human review process for sensitive outputs.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A startup is choosing between two LLMs for a chatbot: one has 7 billion parameters and the other has 70 billion parameters. They have limited GPU resources and need fast response times. They also need the chatbot to handle complex multi-turn conversations. Which consideration is most important when making this choice?

   • AThe number of parameters is irrelevant; only the training data matters for quality.
   • BThe smaller model will always produce lower-quality responses, so the larger model is always the better choice.
   • CThe larger model is easier to fine-tune because it has more parameters.
   • DThe larger model may provide better quality but requires significantly more computational resources, which could slow response times and increase costs.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is 'hallucination' in the context of LLMs?

   • AThe model repeats the same response for different inputs
   • BThe model's training data contains errors
   • CThe model refuses to answer a question
   • DThe model generates plausible-sounding but factually incorrect information
   Reveal Solution Discussion Select an answer first