Why ChatGPT Makes Things Up — and the Fix Nobody Told You | Paragon AI MAG

Why ChatGPT makes things up — and the fix nobody told you. ChatGPT invented six court cases that never existed, real lawyers filed them in a real court, and got fined $5,000. This Paragon AI MAG explainer shows why language models invent facts (they predict the next word — they do not look facts up), the two kinds of made-up answers, the real Mata v. Avianca case, and the fix: retrieval-augmented generation (RAG) — make AI look things up first, then check the page it points to. You also get a 3-step walkthrough you can use tonight and a 5-rule check-before-you-trust list for money, health, legal and school decisions. RAG reduces made-up answers; it never eliminates them, and this video says so honestly. Subscribe to Paragon AI MAG for one video and one podcast episode every day, every fact checked against sources several times before release. Full video + podcast links: Full video:    • Why ChatGPT Makes Things Up — and the Fix ...   · Podcast episode: https://redcircle.com/shows/cbce814a-d451-... · Article: https://paragonaimag.substack.com/p/why-ch... Follow Paragon AI MAG: YouTube    / @paragonaimag   · Instagram https://www.instagram.com/paragonaimag/ · TikTok https://www.tiktok.com/@paragonaimag · X https://x.com/ParagonAIMag · Threads https://www.threads.com/@paragonaimag · Bluesky https://bsky.app/profile/paragonaimag.bsky... · Substack https://substack.com/@paragonaimag · Medium https://medium.com/@paragonaimag · Dev.to https://dev.to/paragonaimag · Pinterest https://www.pinterest.com/paragonaimag/ · Flipboard https://flipboard.com/@paragonaimag · Paragraph https://paragraph.com/@paragonaimag Chapters: 00:00 Hook: six invented court cases, $5,000 fine 00:59 Roadmap 01:19 Part 1: why models invent facts 02:00 Two kinds of made-up answers 02:28 The scale: hallucination rates 03:06 Part 2: the $5,000 legal disaster — Mata v. Avianca 04:12 The scale explodes: 1,300+ cases, and voter-guide tests 04:42 Part 3: RAG, the open-book fix 05:53 Part 4: how AI searches files, step by step 07:15 Part 5: why RAG still fails 08:36 Part 6: rules for trusting AI 10:02 Payoff: give it the book, check the page.