🕵️ Hallucination Detector Game

Learn to spot AI hallucinations in text, images, and audio

0Total Score
0Games Played
0%Accuracy
1Level

🎓 Understanding AI Hallucinations

Text Hallucinations

AI models can generate convincing but false information. Common patterns include:

  • Fake Facts: Incorrect dates, statistics, or historical events
  • Invented Citations: Non-existent research papers or authors
  • Overly Specific Claims: Precise statistics without credible sources
  • Extraordinary Claims: Dramatic statements without verification

Example Hallucination:

"The Great Wall of China was built in 1969 and can be seen from the Moon with the naked eye."

❌ Wrong date (built over centuries) and false claim about Moon visibility

Image Hallucinations

AI-generated images often contain telltale artifacts and impossibilities:

  • Anatomical Errors: Extra limbs, wrong number of fingers
  • Physics Violations: Impossible perspectives, uphill water flow
  • Inconsistent Lighting: Multiple light sources, wrong shadows
  • Background Artifacts: Nonsensical objects or text

Common Issues:

🖐️ People with 6+ fingers
💡 Inconsistent lighting/shadows
🏗️ Impossible architecture

Audio Hallucinations

Synthetic audio has distinctive characteristics that betray artificial generation:

  • Unnatural Rhythm and Intonation: Robotic speech patterns, weird emphasis
  • Voice Inconsistencies: Sudden pitch changes, modulation errors
  • Background Anomalies: Inconsistent ambient sound
  • Phantom Sounds: Voices or words that weren't actually spoken

Red Flags:

🎵 Sudden background music changes
🗣️ Robotic voice modulation
👻 Phantom voices

Detection Tips

Glossary of Key Terms

  • AI Hallucination: When an AI confidently gives you an answer that sounds real but is made up. Example: “The Moon is made of cheese.”
  • Prompt: What you type or say to ask something from the AI. Example: “Tell me a joke about robots.”
  • Metadata: Info about a file, like who made it, when it was made, or what it’s about. Example: A photo’s date and camera settings.
  • Deepfake: A video or voice that looks real but was created using AI. Example: A fake video of a celebrity saying things they never said.
  • Bias: When AI gives unfair or one-sided answers because of what it was trained on. Example: Only showing male CEOs in search results.

🎯 Choose Your Challenge

📝

Text Detection

Identify fake facts and misinformation in AI-generated text

Best: 0%
🖼️

Image Detection

Spot anatomical errors and impossible artifacts in images

Best: 0%
🎵

Audio Detection

Recognize synthetic speech and audio anomalies

Best: 0%

🏆 Your Achievements

Text Detection
Question 1 of 5Score: 0

🎉 Game Complete!

0Final Score
0%Accuracy