Today you interview an AI chatbot — not to use it, but to study it. The assignment is symmetrical and strange: come back with three things it does astonishingly well, and three ways you made it confidently wrong. Both halves are required, because the machine’s strangest property is that its brilliance and its nonsense arrive in exactly the same fluent, assured voice.

The task

In pairs, run a structured interview and keep a lab record: every prompt you tried, what came back, and your verdict. For the three strengths, push until you are genuinely surprised. For the three failures, “I don’t know” does not count — you are hunting for answers that are wrong and confident at once, and you must be able to show the class evidence of why the answer is wrong.

What tends to surface

The tone never changes. Right or wrong, the answer arrives polished, confident, and grammatical — which means fluency is no evidence of truth. That asymmetry sets the terms for using these tools at all: wonderful where you can check the output, hazardous where you cannot. Automation and Artificial Intelligence digs into why, and what to do about it fuels Will AI Take the Jobs.

An extension

Swap failure lists with another pair and try to reproduce their results. Some failures vanish on the second try — which raises the harder question of what it means to test a machine that does not answer the same way twice.

The answer is not on this page

No verdict on AI is printed here. The class assembles its own picture from the evidence every pair brings back — and the picture will not match last year’s, because the machines will have changed.

Curriculum connection

B4.1

investigate current innovations, including automation and artificial intelligence systems, and assess the impacts of these technologies on everyday life

Link to original