Nearby lessons

13 of 30

Hallucination

Concept in One Line

Hallucination means GPT gives an answer that sounds correct and confident, but is actually wrong or not verified.

Why This Concept Is Important

  • It prevents blind trust in AI answers.
  • It is very important for exams, office work, and projects.
  • It is frequently asked in interviews.

Key Points to Remember

  • GPT predicts text, it does not verify facts.
  • Confident answer does not mean correct answer.
  • Hallucination happens when data is missing or unclear.
  • GPT may guess to complete the answer.
  • Human verification is always needed.

Why Hallucination Happens

Reason 1: Prediction, Not Checking

  • GPT predicts the next word.
  • It does not automatically check facts.

Reason 2: Missing or Unclear Data

  • Exact information is not available.
  • GPT fills gaps with the best guess.

Reason 3: Confident Language Style

  • GPT is trained to sound helpful.
  • Confidence can hide uncertainty.

Reason 4: Vague Questions

  • Unclear prompt -> higher chance of wrong answer.

Simple Daily Life Example

  • A friend gives wrong train timing confidently.
  • It sounds sure, but the information is wrong.

Same thing happens with AI.

Simple Flow

Unclear Question / Missing Data -> Guessing -> Confident Output -> Possible Wrong Answer

Interview / Exam Points

Q1: What is hallucination in AI?

Hallucination is when an AI gives a confident but incorrect answer because it predicts text instead of verifying facts.

Q2: Is hallucination a bug?

No. It is a limitation of how language models work.

Common Confusions to Avoid

  • AI is not always correct.
  • Confidence does not equal accuracy.
  • AI answers must be verified.
  • Use AI as an assistant, not a final authority.

One Line to Remember

Hallucination = Confident answer, but wrong fact

Quick Self-Check

  • GPT always gives correct facts -> Yes / No?
  • Hallucination means confident but wrong answer -> Yes / No?
  • Correct answers: No, Yes

🧠 Test Your Knowledge

3 Questions

Progress: 0 / 3
Keep Going!GPT Strengths and Limitations