가이드 연계 퀴즈 · 쉬움 레벨

AI 시스템 사고 퀴즈

전체 워크플로에서 AI 서비스 입력, 종속성, 피드백, 병목 현상, 경계 오류, 엔드투엔드 결과 및 버전 기록을 매핑합니다.

관련 가이드 경로AI 시스템 사고
질문 1 의 8

In an AI service, what makes a feedback loop?

모든 질문 미리보기8 질문

퀴즈를 풀면 정답과 설명이 나타납니다.

  1. In an AI service, what makes a feedback loop?
    • A system output changes data or behavior that later returns as input
    • A model is saved to a backup drive
    • Two developers review the same code
    • A user closes a browser tab
  2. Which sequence best maps the complete path described in the guide?
    • Model, then interface only
    • Input collection, preprocessing, retrieval, model, tools or review, storage, and feedback
    • Training data, then model weights only
    • User account, billing, and logout
  3. Why should a team record the owner and failure behavior of each dependency?
    • To increase the model’s parameter count
    • To remove the need for monitoring
    • To know who is responsible and what happens when a component or boundary fails
    • To ensure every component uses the same programming language
  4. A workflow spends 40 seconds in review while retrieval and generation take 3 seconds combined. Where should a systems team investigate first?
    • 모델의 훈련 날짜
    • The logo shown in the interface
    • The number of model layers
    • The review bottleneck and how source presentation could improve it
  5. Why can recommendation exposure affect later system behavior?
    • User reactions to shown items can become future data
    • The interface changes the model’s source code automatically
    • A recommendation always reveals a user’s identity
    • Exposure prevents future model updates
  6. Which issue can create a wrong outcome without a model crash?
    • A clear ownership record
    • A missing field or outdated cache at a component boundary
    • A version number in a log
    • A reviewed source document
  7. What should an end-to-end check verify?
    • Only that a model endpoint returns a response
    • Only that the final screen loads
    • The user-visible result across the complete workflow
    • Only the number of retrieval documents
  8. What information helps trace a regression after a system change?
    • Only a screenshot of the home page
    • 모든 사용자 목록
    • A model’s marketing name
    • Version information for the components involved