I DISAGREE WITH FRANCIS: Pangram and Substack, why the whining? My suspicious brain thinks it could be that they don’t want to be found out.
What do I disagree with? Well… first, that using AI detection tools is a good thing. In my experience, including various experiments I’ve run, they are about as good as early COVID tests, and therefore I discourage them for the same reason. If your results are as good as flipping a coin, and if people are going to do the stupid in answer to the test: DON’T DO IT. (Not that I’m going to disable it in my substacks. Couldn’t be bothered.)
Second, I disagree with him on the “tells” of AI. I have found a lot of tells myself (Look, my guilty pleasure if JAFF which at this point is almost all AI written) including an unhinged love for all forms of the verb “perform” used in the most bizarre places and times. Also a certain confusion about how many parents go into making a human which leads to sentences such as “inherited her eyes from one of her fathers.” This stuff seems to crop up consistently, as does an inability to understand physicality or distance. BUT sentence structure and cadence is mostly going to ping all the writers that AI was trained on. Writers like me whom Anthropic wholesale used as training material. And seriously, people, I was abusing the English language long before AI was a twinkle in a server’s eye. My bizarre tricks of construction are very much the mark of a human being. Also if we’re going to chase sentence structure, there is a lot of room to find witches everywhere.
Mostly I’m tired of all the nonsense. There’s good and there’s bad. There’s entertaining or not. How it’s produced is a matter of complete indifference to me.
Only if you’re using a clanker to write JAFF, can you please for the sake of all that’s holy edit out all the performing? The rest is sometimes hilarious, but the performing is just tedious.
