The Guardian has published an opinion piece it says was written by its new language generator, GPT-3.
And yes, they subbed it before publication, cut lines and paragraphs, and sometimes swapping them around, but overall they say, "it took less time than many human op-eds".
So do journalists have anything to fear from robots? In a print-publishing world, there might be competition for space, but in this digital one, the competition will merely be for attention. And GPT-3 says he/she/it will just sit in the background, while humans "do their thing... hating and fighting each other".
The Guardian says the robot was told to write around 500 words, explaining why humans have nothing to fear from AI, and was fed an introduction.
After being fed prompts, written by the Guardian and fed by UC Berkeley computer science undergraduate student Liam Porr, GPT-3 produced eight veersions. "Each was unique, interesting and advanced a different argument," the publisher says. Rather than running any one, it has published "the best parts of each, in order to capture the different styles and registers of the AI".
Comments