Comment by mapontosevenths
18 hours ago
My first instincts offhand were:
* N-gram analysis of lyrics. Even good LLM's still exhibit some weird pattern's when analyzed at the n-gram level.
* Entropy - Something like KL divergence maybe? There are a lot of ways to calculate entropy that can be informative. I would expect human music to display higher entropy.
* Plain old FFT. I suspect you'll find weird statistical anomalies.
* Fancy waveform analysis tricks. AI's tend to do it in "chunks" I would expect the waveforms to have steeper/higher impulses and strange gaps. This probably explains why they still sound "off" to hifi fans.
* SNR analysis - Maybe a repeat of one of the above, but worth expanding on. The actual information density of the channel will be different because diffusion is basically compression.
* Subsampling and comparing to a known library. It's likely that you can identify substantial chunks that are sampled from other sources without modification - Harder because you need a library. Basically just Shazam.
* Consistency checks. Are all of the same note/instrument pairs actually generated by the same instrument throughout, or subtly different. Most humans won't notice, but it's probably easy to detect that it drifts (if it does).
That's just offhand though. I would need to experiment to see which if any actually work. I'm sure there are lots more ways.
Thank you!
This will likely have a lot of false positives on a lot of genres. E.g. I suspect genres like synthpop and trance (and a lot of other electronic music) will likely hit a lot of those points with regards to music and sampling.
Lyrics are also not a given (when they are likely curated by humans). E.g. compare the song I referenced (https://dumpstergrooves.bandcamp.com/track/he-talked-a-big-g...) to, say, Taylor Swift's current most listened to song: https://genius.com/Taylor-swift-the-fate-of-ophelia-lyrics I'd chose the AI one in a heart beat :)
I wonder if a combination of all of those may work for a subset of songs, but I don't think you can do it with any confidence :(
> I wonder if a combination of all of those may work for a subset of songs, but I don't think you can do it with any confidence :(
Thats a solid point. Pretty much all of my ideas are probabilistic. I suspect you're right and it will have to work a bit like spam detection, where each "fail" for a test is seen as one indicator that adds to a score. Then above a threshold score it's flagged for further review and sent to a "spam" folder where a human can judge.