> If they do, then AI writing will incorporate them.
Eventually, but at least for something like "ain't" the hill to climb to get there will probably be a lot higher, because you've got decades if not centuries of explicit discouragement and very little presence in existing formal writing (and maybe online writing period).
The more durable proof of humanity will probably be hyper rapid slang evolution, but that's going to have the downside of social fragmentation and making the past less legible. If slang changes rapidly over the course of a year, will anyone still remember the slang from 10 years ago to understand an old post?
One thing I have yet to see on the mainstream AI platforms is a combination of foul language, dirty jokes, rule34 and racial slurs. 4chan-GPT and local LLM's excluded. Perhaps require Rule34 and foul language as proof. There's probably a way to jailbreak that for a little bit but it would get patched.
Eventually, but at least for something like "ain't" the hill to climb to get there will probably be a lot higher, because you've got decades if not centuries of explicit discouragement and very little presence in existing formal writing (and maybe online writing period).
The more durable proof of humanity will probably be hyper rapid slang evolution, but that's going to have the downside of social fragmentation and making the past less legible. If slang changes rapidly over the course of a year, will anyone still remember the slang from 10 years ago to understand an old post?