At the time of market creation, AI music models were already able to generate songs of 80 seconds in length that approached the quality of what an average musician could output. How long will it be until AI becomes superintelligent at music?
This market will resolve to YES immediately if a mainstream media outlet publishes an article stating that an AI model has produced a work that is superior to that written by Bach, Beethoven, or Mozart before December 31, 2026. The work may be from any genre and may be a type of audio waveform that was not previously known to humans, or which is not possible to create with musical instruments.
The waveform must be original and may not simply be a superior remix of one of those composers' works. It will also resolve to YES if the general public consensus is that AI music is now better than human music in all respects. A YES resolution will not be reversed if there is controversy over the article.
The market will resolve to NO on the close date otherwise.
The resolution criteria are completely useless to measuring AI progress.
It's entirely possible for some journalist to call some mediocre piece "better than Mozart" for clicks.
It's even more likely than the next masterpiece won't be recognized as such until it stands the test of time. Classical music gets so much praise not because it's just good, but because it's still good centuries later.
I don't understand at all why this market is so low.
If I could bet on it, I would take odds of 80%. Udio has almost achieved this already, today. The sonatas it is putting out are indistinguishable from the best musicians when considered as individual verses. As soon as it can understand how to develop the piece as a whole, it will already be superintelligent.
I expect this to resolve to YES within 6 months, at the latest.
@SteveSokolowski Udio is impressive, and its classical music is indistinguishable to me (someone with no background in classical music), but "clearly better than Mozart" is a very high bar. I'd expect it to continue to improve over the next couple years, and I doubt I'll be able to tell apart AI-generated and human-created music by EOY 2026, but that doesn't mean no one will be able to. I'm sure there are complicated things Mozart was doing in his music that Udio does not do.
@SteveSokolowski I don't think that superhuman-in-quality artworks exist for text or images, either—the problem is eliciting those capabilities from generative AI models, not that they're not present.
(Though in the case of music my best guess is that they're not present.)
@ArmandodiMatteo "this one minute thirty mumble rap wav is sicker then anything Mozart could have made"