Recent Advancements and Future Directions in AI AIE Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A startup is building an AI that generates legal contracts. The AI must produce legally sound documents, but the startup has limited compute resources and cannot afford to fine-tune a massive model. They also need to ensure the AI does not generate clauses that contradict each other. Which strategy is most effective given these constraints?

   • AUse a pre-trained language model with a constrained decoding algorithm that enforces legal templates and consistency rules.
   • BUse a fully autonomous system that generates contracts without any oversight.
   • CUse a generative adversarial network to create contracts from scratch.
   • DFine-tune a large language model on a corpus of legal documents despite the cost.
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A research lab is developing an AI that can write scientific papers. They want the AI to generate novel hypotheses but also ensure the content is accurate and reproducible. Which approach best balances creativity and reliability?

   • AUse a model that only rephrases existing papers
   • BAllow the AI to publish without any verification
   • CUse a generative model with a validation step that checks claims against known databases
   • DUse a purely rule-based system to generate papers
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A company wants to use AI to automate customer service. They have a large volume of inquiries but also a high bar for customer satisfaction. They are considering a fully automated chatbot versus a human-assisted system. Which factor is most important in this decision?

   • AThe complexity and emotional sensitivity of customer inquiries
   • BThe speed of the AI model's response
   • CThe cost of the AI infrastructure
   • DThe number of languages the chatbot supports
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A national bank wants to use AI to detect fraudulent transactions in real time. The system must minimize false positives to avoid blocking legitimate customers, but also catch as much fraud as possible. The bank also needs to comply with anti-money-laundering regulations that require reporting suspicious activity. Which approach best balances these competing goals?

   • ASet the model to flag all transactions that are even slightly unusual to maximize fraud detection.
   • BAutomatically block all flagged transactions without human review to ensure compliance.
   • CUse a machine learning model with a tunable threshold, and route borderline cases to human investigators for review.
   • DUse a simple rule-based system that only flags known fraud patterns.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A retail chain is implementing AI-powered demand forecasting to optimize inventory. They notice that the model performs well on historical data but fails during unexpected events like a pandemic. What is the most likely reason for this failure?

   • AThe model is too small to process all transactions
   • BThe model cannot generalize to patterns outside its training distribution
   • CThe training data was not stored in the cloud
   • DThe model uses too many features
   Reveal Solution Discussion Select an answer first