tech

AI systems like ChatGPT or Claude can confidently present nonexistent papers, events, or people as real facts — a failure known as 'hallucination' that never fully disappears even with more training data.

What is its structural cause?

Because models are trained to predict the most statistically likely next word, with no separate mechanism to verify whether a statement is actually true ✓
Because the total volume of training data is still insufficient, causing the model to fill gaps with fabricated content
Because the model has no real-time internet connection and therefore lacks access to current information

Because models are trained to predict the most statistically likely next word, with no separate mechanism to verify whether a statement is actually true
An LLM is a probability engine that generates plausible-sounding text — it is not a system that queries and cross-references a factual database.

▶ Quration Play

tech

What is this Microsoft chip called?What is this model called?Which was invented first — the fax machine or the telephone?Who is widely regarded as the world's first computer programmer?What does the "http" at the start of a web address stand for?What does "Wi-Fi" actually stand for?What was that first website about?How many satellites does a GPS device need to pinpoint your location?What was the very first message sent over the early internet in 1969?Junk email is called "spam" because of what?The first computer "bug" was literally what?What is a "cookie" in web browsing?

Quration — Quration Play