引導式測驗 · 中等 等級

藝術類測驗中的人工智慧影像產生器

測試擴散模型如何製作圖像、黎明查莉亞版權決策以及藝術課的歸屬和安全實踐。

問題 1 的 8

擴散模型如何從提示生成圖像?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. 擴散模型如何從提示生成圖像?
    • It starts from random noise and refines it step by step into an image that matches the text
    • It searches the web and returns the closest photo
    • It cuts and pastes pieces of stored images together
    • It traces over an image the user uploads
  2. 美國版權局於 2023 年對『Zarya of the Dawn』作何決定?
    • The whole comic, including the images, was protected
    • Nothing in the comic could be protected
    • The text and arrangement were protected, but the individual Midjourney images were not
    • Only the Midjourney images were protected
  3. Adobe 說 Firefly 訓練於哪些內容?
    • Only images scraped from social media
    • Licensed content such as Adobe Stock, plus public domain material
    • Students' uploaded artwork
    • Nothing; it has no training data
  4. 隨機種子在影像生成中控制什麼?
    • The colour palette
    • 圖像的檔案大小
    • Which tool is used
    • The starting noise, so the same prompt and seed usually produce the same image again
  5. 為什麼指南說圖像對圖像模式往往適合教室?
    • They keep more of the student's own composition
    • They are always free
    • They never show bias
    • They need no prompt at all
  6. Which statement corrects a common misconception about image generators?
    • They store every training image and paste them together
    • They cannot produce anything resembling their training data
    • They learn statistical patterns, though they can sometimes produce close copies of images that appeared many times in training
    • They only produce abstract art
  7. What attribution practice does the guide recommend for AI-assisted student work?
    • Hide AI use to keep grading fair
    • Label the work, name the tool, keep the prompts, and describe what the student changed
    • Credit only the AI company
    • Attribution is only needed for professional artists
  8. What does the guidance scale setting control?
    • The age limit for users
    • How many images are produced at once
    • The image's resolution
    • How closely the image follows the prompt versus how natural it looks