That makes sense as an issue for polished, popular writing that has the goal of reaching many new readers by being easy to read and about a topic that’s not very specialized. Is it an issue for more niche things?
i think it’s an issue everywhere. the only time it’s not a problem is if i trust the writer enough that i know it’s not AI written even if it looks to be; or they would be thorough even if it was AI written.
no, because it doesn’t feel like LLM text. em dashes are neither necessary nor sufficient to make a text feel like LLM text. but they’re one feature in the classifier. if this text also had lots of “it’s not X, it’s Y” and so on, then I’d be more suspicious.
That makes sense as an issue for polished, popular writing that has the goal of reaching many new readers by being easy to read and about a topic that’s not very specialized. Is it an issue for more niche things?
i think it’s an issue everywhere. the only time it’s not a problem is if i trust the writer enough that i know it’s not AI written even if it looks to be; or they would be thorough even if it was AI written.
I mean, suppose for example you read the abstract of this: https://berkeleygenomics.org/articles/Chromosome_identification_methods.html
It has an em dash, and it is in a dry / boring / formal / academic / passive tone. But are you really worried it might be LLM text?
no, because it doesn’t feel like LLM text. em dashes are neither necessary nor sufficient to make a text feel like LLM text. but they’re one feature in the classifier. if this text also had lots of “it’s not X, it’s Y” and so on, then I’d be more suspicious.