flâneur

Why do AI models use so many em-dashes?

seangoedecke.com · 2,133 words · saved by 1 readers

If you asked most people to name a defining feature of AI-generated writing, they’d probably say the em-dash — like this. Language models use em-dashes so much that real humans who like em-dashes have stopped using them out of fear of being confused with AI. It’s also surprisingly hard to prompt models to avoid em-dashes: take this thread from the OpenAI forums where users share their unsuccessful attempts. Given all that, it’s kind of weird that we don’t really know why language models use the em-dash so much. One common explanation is that normal English text contains a lot of em-dashes, so it’s just learned behavior from the training data. I find this fairly unconvincing, for the reason that everyone thinks AI uses a lot of em-dashes. If em-dashes were as common in AI prose as human prose, they would be as unremarkable as the use of other punctuation marks. Another explanation I’m not convinced by is that AI models like em-dashes because they’re so versatile. When the model is tryin

If you asked most people to name a defining feature of AI-generated writing, they’d probably say the em-dash — like this. Language models use em-dashes so much that real humans who like em-dashes have stopped using them out of fear of being confused with AI. It’s also surprisingly hard to prompt models to avoid em-dashes: take this thread from the OpenAI forums where users share their unsuccessful attempts. Given all that, it’s kind of weird that we don’t really know why language models use the em-dash so much . Explanations I don’t find convincing One common explanation is that normal English

related reading