JA Westenberg

The AI-based AI detection tools are just playing an eternal // infernal game of whackamole with this model and that model and the next model.

Snake, meet tail.

You're going to get along swimmingly.

https://www.joanwestenberg.com/the-ai-writing-witchhunt-is-pointless/

The AI writing witchhunt is pointless.

Alexandre Dumas ran what was essentially a content production house in 19th century Paris. His most famous collaborator was Auguste Maquet, who wrote substantial portions of The Three Musketeers and The Count of Monte Cristo. Maquet would produce drafts and outlines, and Dumas would rewrite and polish them, but the

Westenberg.
12:02pmTypefully
Show threadNiels Abildgaard12h ago

@Daojoan I have been thinking similar things, maybe framed a bit more broadly here: https://mas.to/@nielsa/116255178545615673

Would love your thoughts on that.

Niels Abildgaard (@[email protected])

As we weather the storm of LLMs move through software development culture, we have to protect the commons, so there is an alternative for when the cultural bubble bursts. This is a good example of rallying around the commons. I wrote more on the general idea here: https://mas.to/@nielsa/116237619916131781

mas.to
Show threadKen Milmore12h ago
@Daojoan It's quite reminiscent of 2000s era Windows security software. We make the holes, and also the inadequate tools to block them up.
Show threadAunty12h ago

@Daojoan the but which gets me, is that people always treat bad writing as evidence of AI, even though AI prose tends not to be particularly bad (while still being generic and boring -- but not bad)

I've seen a similar dynamic with visual art. People will point at awkwardly drawn hands and declare it AI slop, (even though hands are notorious as the but amateur artists struggke with)

Show threadAndrew (Television Executive)11h ago

@Daojoan this is great, and speaks directly to fears and anxieties I've experienced first hand.

I find myself re-reading and re-editing things I wrote years ago because suddenly "that sounds a little LLM-y"

Purity test and witch hunts will almost always hurt the people they are allegedly defending.

Show threadKrypt3ia11h ago
@Daojoan #ouroboros
Show threadmyrmepropagandist11h ago

@Daojoan

I agree. Detection tools are kind of useless. If the writing is bad it's bad. One way writing can be bad is when it "sounds like AI" ... we can break down what we mean by that, but bad is bad.

Show threadNemo11h ago

@Daojoan

"He stood up, leaving the chair on which he was sat [1]" is genuine Dumas (my translation). The man was a master of padding, being paid by the line.

[1] Il se leva de la chaise où il était assis.

Show threadBlender Dumbass ( J.Y.Amihud )10h ago
@Daojoan What we need is not AI tests. We need a guilty until proven innocent mindset. If the writer doesn't produce a screen recording of how they typed the text, it's AI.