The other day, conversation turned to love. Birthday cards in particular. What the hell are you meant to write in these things? Happy birthday big man. Love you lots. It’s hard to squeeze meaningful sentiment into a paper flap, we all agreed. “My husband used ChatGPT,” someone said. ChatGPT? Oh yes, pop in a prompt and it will tell you what to say.
I spoke to the husband in question, Jan. “I just thought I’d try it out, likely even with a dose of scepticism to see if it would do a decent job,” he said. “My reaction to the first attempt was ‘not bad at all’. I then gave it three or four more prompts to improve and was pretty pleased with the final version.” He wrote down the results, licked the envelope and handed it to his friend.
During the past few years, OpenAI’s ChatGPT and other large language models have entered society with increasingly loud footsteps. These chatbots repackage information drawn from enormous data sets of books, websites, social media posts, software manual pages and a long list of other sources, parsing it into grammatical text. ChatGPT alone has an estimated 800mn weekly active users as of May 2025, a staggering twofold raise from an estimated 400mn in only February 2025.