Practical exercise
Catch AI being wrong on purpose
Trusting AI without checking is like signing a blank cheque. Today you will run three small traps, designed to catch AI when it is confidently wrong, and write down where it broke. After this you will never again read AI output without a small voice asking, is this actually true.
25 min·starter
Now try it
Catch AI being wrong on purpose
Stretch
If you have ChatGPT free tier access on a laptop, ask the same three questions to GPT and to Google's Gemini free tier in different tabs. Note which one fabricated more, and which one admitted it did not know. Save the screenshots side by side.
Goes into your portfolio
A three column trust audit on paper showing question, AI answer, real answer, with green ticks and red crosses, plus your personal trust rule.