Artificial Intelligence · 人工智慧單元測驗卷

一、單字配對（20 分）

把右邊的代號填進左邊的括號裡。

二、選擇題（30 分）

1. (    )1.How does machine learning differ from a traditional program?

   1. (A) It is faster
   2. (B) The system finds the rules from examples instead of a person writing them
   3. (C) It needs no computer
   4. (D) It never makes mistakes

2. (    )2.What is bias in a model?

   1. (A) A speed problem
   2. (B) Answers that lean the same way the training examples did
   3. (C) A kind of bug
   4. (D) A storage limit

3. (    )3.What does a language model actually produce?

   1. (A) Verified facts
   2. (B) The words that most plausibly come next
   3. (C) A database lookup
   4. (D) A translation only

4. (    )4.What is a hallucination?

   1. (A) A slow response
   2. (B) An invented fact, source or quotation
   3. (C) A picture
   4. (D) A hardware fault

5. (    )5.Which task suits a traditional program better?

   1. (A) Recognising cats in photos
   2. (B) Calculating wages
   3. (C) Writing poetry
   4. (D) Understanding speech

6. (    )6.Why ask a model to show its reasoning?

   1. (A) It answers faster
   2. (B) Mistakes become much easier to spot
   3. (C) It uses less memory
   4. (D) It is required by law

三、填空（每題填一個單字）（30 分）

1. 1.The system finds the ____ in the examples by itself.（提示：規律）
2. 2.Unbalanced training data leads to ____.（提示：偏誤）
3. 3.A language model ____ the words that come next.（提示：預測（第三人稱））
4. 4.An answer can be ____ and still be wrong.（提示：看似合理的）
5. 5.Always ____ anything that matters.（提示：查證）
6. 6.Asking for its ____ makes mistakes easier to see.（提示：推理）

四、簡答（用英文回答，一到兩句即可）（20 分）

1. 1.Why can an AI system be confidently wrong?為什麼 AI 會非常有自信地講錯？

2. 2.Explain why the training data matters so much.說明訓練資料為什麼這麼重要。