The Hard Part: Plausibly wrong? What a Year 8 history lesson teaches about AI hallucinations

You probably sat through it aged 12: a history lesson about sources. Who wrote this, why and what are they not telling you. Turns out that lesson is super useful for working with AI. In this short episode, Helen Dawson ( child of a history teacher) draws the line straight from GCSE source analysis to how large language models actually work. Generative AI has been fed a vast, partial, skewed corpus of everything ever written — and it answers by predicting the next word. Which means it's never truly right. It's only ever plausibly wrong, and sometimes that happens to line up with the truth. There's no signpost telling you which. Helen makes the case that critical thinking is the number one skill of our time, and walks through the questions that protect you: Is this even the right tool for the job? What was this thing trained on? Did the way I asked the question seed the answer I got? And why footnotes, however much they made you cry during your degree, still matter. (If you weren't an Arts student, just trust us on this) Guardrails have improved. They are not hard limits. Check your sources.