Public and media interest in AI has taken off dramatically in the last few months: when a friend showed me ChatGPT just before Christmas 2022 it was a novelty, and we spent an evening asking it to write increasingly outlandish stories about our friends. At the time of writing, public opinion appears to be that AI language models such as OpenAI’s GPT-4 are rapidly advancing and may soon be able to take over the jobs of us feckless humans. At the same time, educational institutions and governments are concerned that students could produce competent work for assessment with little or no effort; plagiarism checkers are rushing to develop tools that can check whether text was written by AI, while developers work on language models to fool those checkers.
