AI and the English Language
George Orwell closed his 1946 essay “Politics and the English Language” with six rules for writers, the first of which commands: “Never use a metaphor, simile or other figure of speech which you are used to seeing in print.” It is the rule against clichés—don’t beat dead horses, cry over spilled milk or freeze like a deer in the headlights.
“The fight against bad English,” Orwell wrote, “is not frivolous and is not the exclusive concern of professional writers.” Authors and academics are often the worst abusers of ready-made phrases and buzzwords. They put their readers to sleep by deploying trite phrases in the passive voice. Good, honest writing isn’t only hard work; it can get you canceled for being direct about a controversial topic.
Enter artificial intelligence, which can flag clichés by analyzing the frequency at which figures of speech occur in data sets. But replacing clichés with something vivid and original is a harder task. In Orwell’s parlance, large-language models show us what they are used to seeing in print. These models are like giant looms. They select complementary textual threads from vast data warehouses and weave them together to create tapestries of prose.
