🤖 AI & AGI Academy · How AI Actually Works · Lesson 3 of 9
Why AI makes things up
Hallucinations and how to catch them
In short: AI predicts the next likely word based on patterns, not facts it checks. When it lacks real knowledge, it still produces a smooth, confident-sounding answer that can be wrong.
How it works
• Remember AI generates text by predicting likely words, with no built-in fact-checker or sense of truth.
• Watch for warning signs: oddly specific names, dates, numbers, citations, or quotes that sound too neat.
• Ask the AI to show sources or explain its reasoning, then check those sources yourself independently.
• Cross-check anything important (names, prices, laws, health or legal facts) with a trusted outside source.
🧮 Example: Amara asks an AI assistant for the population of her town in 2023 and a quote from its mayor. It gives a precise number and a confident quote, both invented. She checks the town's official website and finds no such data exists.
✅ Try this today: Ask any AI assistant a question about a specific fact you can verify, then spend 5 minutes checking its answer against a trusted source online.
⚠️ Common mistake: Trusting confident, detailed answers as automatically true. Always verify names, numbers, quotes, and dates before relying on them.
➡️ Next lesson: Use AI at Work: Research without getting fooled
🔖 Save this — the quiz is below 👇
News | Chrono | 𝕏 takes