It's really easy to have a false positive and false positives can be very harmful if the person using the detector isn't aware of that risk.

It's also very easy to change the pattern of LLM output. You can provide basic prompting that will significantly change the structure of the output. For example, having it utilize the Wikipedia article on signs of AI writing and avoid everything it describes. https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing

But also their algorithmic playlists have gotten worse and worse. They overfit user data. They all recommend the same singles over and over and over. I've found they also don't make sense, recommending music that doesn't even belong in the playlist.

I've switched to user created playlists more often than the algo ones, although a lot of user playlists are just saved copies of Spotify generated recommendations.

That is an SUV, not a wagon.

As others have pointed out elsewhere this is literally the type of behavior OpenAI is founded on. Gathering up other people's IP and using it to build their own thing. It's how all the big LLMs are built.

New cars with intrusive driver monitoring alerts are obviously going to be terrible but you can still buy vehicles made prior to this change.

> "...researchers demonstrate how Vantablack® 310 – an ultra-black coating developed by University of Surrey spinout Surrey NanoSystems, co-authors of the paper..."