A lot of music content shows up in my feed these days. I have been exploring music—theory, performance, recording and even composition—as a creative side-project for a couple of years, (often writing about it over on my other blog.)

Musicians, like many other creative professionals, are struggling with the emergence of generative AI systems and many of the ones I watch are pushing through the noise of the AI-hype engine and just trying to keep making music.

One channel, though, snagged in the overgrowth of my consciousness recently when during a nine minute video debunking a deepfake equipment review, Youtuber Rick Beato^[1] noted, almost as a side thought, that the AI voice algorithm was struggling with a very specific word: bass.

If you are a music person, bass is pronounced with a long A, like: Baaaay

If you are a fish person, bass is pronounced with a short A, like: Baaaah

The AI was pronouncing it with a short A, like the fish.

Oh, I’m sure plenty of humans make the same mistake. But those humans should not have any credibility and certainly not wide followings on music YouTube, right?

It’s a subtle but pretty easy catch for anyone inside a couple standard deviations of the industry, and it speaks to an interesting story that wraps up the current state of generative AIs, authenticity and the nuances of real human knowledge versus slop.

Mr. Beato didn’t ignore the gap, but he so easily could have turned his whole video into a single quip on the lacking authenticity and collapse of the trust equation that comes when a video, either human made or generated, can’t tell the difference between the instrument it’s hyping and a big old cold fish.