BriefTea
Every story in sixty words
Get the free app no ads · no paywallExplained in plain English
AI models 'hallucinate' because they are designed to predict the next most plausible word or phrase based on their training data, rather than to know facts. If their training data is incomplete, biased, or if the query is ambiguous, the AI can generate confident but incorrect information, as Google admits with Hydrogen. It's a fundamental limitation of current generative AI.
Stories that explain this
Google Search now uses new Gemini 4 Hydrogen model to answer multi-part questionsRelated explainers
How does AI get faster? Why is it called Hydrogen? How much does AI cost? Will AI make things cheaper? How does AI make attacks worse? Will AI make council jobs redundant? Will AI make councils more efficient? How does AI make money?The full catalogue
Browse every card filed under A →We would like to count visits and see which links bring readers here. That means storing a random ID in your browser. No cookies, no ads, no profile. How it works.