Ethical, Societal, and Security Concerns in AI AIE Practice Questions (Page 8)

Questions 36–40

1. 36foundation · easy

   Which security threat is characterized by an attacker making small, often imperceptible changes to input data to cause an AI model to make incorrect predictions?

   • AMembership inference
   • BData poisoning
   • CAdversarial attack
   • DModel theft
   Reveal Solution Discussion Select an answer first
2. 37application · medium

   A healthcare startup develops an AI system that diagnoses skin conditions from photos. The system is trained on a dataset that is predominantly light-skinned patients. The startup wants to expand to a global market and avoid criticism for potential bias. The startup has limited resources and must choose a mitigation strategy. What should the startup do?

   • AAdd a disclaimer that the system may not work well on darker skin tones
   • BCollect additional training data from diverse skin tones and re-evaluate the model's accuracy across subgroups
   • CUse a more complex deep learning model to improve overall accuracy
   • DDeploy the system only in regions with predominantly light-skinned populations
   Reveal Solution Discussion Select an answer first
4. 38expert · hard

   A social media company uses an AI system to recommend content. The system is vulnerable to adversarial manipulation where users can post crafted content to make the recommendation engine promote their content unfairly. The company wants to maintain user trust and ensure fair content distribution. They also need to keep the recommendation system fast and scalable. Which approach is most effective?

   • AIncrease human moderation of all content before it is recommended.
   • BImplement robust training with adversarial examples and use anomaly detection to flag manipulated content.
   • CDisable user-generated content to eliminate the attack vector.
   • DUse a simpler recommendation algorithm that is less prone to manipulation.
   Reveal Solution Discussion Select an answer first
5. 39application · medium

   A financial institution uses an AI model to detect fraudulent transactions. The model is trained on transaction data that includes sensitive customer information. The institution wants to comply with privacy regulations while maintaining the model's effectiveness. What should they do?

   • AEncrypt the training data at rest and in transit.
   • BApply differential privacy to the training data to protect individual records.
   • CUse a smaller training dataset to reduce exposure of personal data.
   • DRemove all personal identifiers from the training data.
   Reveal Solution Discussion Select an answer first
6. 40application · medium

   A large online platform uses an AI recommendation system to maximize user engagement. The system has been criticized for promoting addictive content and spreading misinformation, leading to negative societal impacts. The platform wants to maintain user growth while addressing these concerns. What should the platform do?

   • AModify the recommendation algorithm to include diversity and quality metrics in its objective function
   • BAllow users to turn off personalized recommendations
   • CHire more human content moderators to review flagged content
   • DAdd a warning label to content that the AI identifies as potentially misleading
   Reveal Solution Discussion Select an answer first