...

I've been a little bit obsessed with OpenAI Jukebox generated songs for the past few months so I've been frequenting a channel on Youtube called Broccaloo that posts all kinds of oddities that the AI randomly spits out. So far it seems like the complete synthesis of musical recording is mirroring the process of the development of modular synthesis into fully reliable synth instruments and it's honestly fascinating to me that I live within the time period of this still being a very niche interest. Sooner or later I'm certain that there's going to be an AI assisted music project from this generation's Terry Riley which will stand the test of time as a work that's ahead of its time. The fact that there's no telling whether or not that project is being crafted right under my nose somewhere around the globe right now is honestly fascinating to me. An entire genre of music could be birthed out of nowhere which is likely going to be lampooned as 'the death of culture as we know it' before people chill out and realize they can still go to local concert venues if they want to be certain they're listening to real people. Maybe it could actually be good for live music...who's to say? 

There's already small elements of AI cooperation in some of my music out there from running samples through a website I've been frequenting called splitter.ai since it's honestly a really reliable tool for isolating vocal or instrumental tracks from mixes. Beforehand where a ridiculous amount of EQ guesswork was required it's just a few button presses to isolate stems from just about anything. Perhaps if someone were a bit of a musical anarchist they could isolate the guitars from some Weezer song then plop some drums from a Phil Collins track in the mix and pitch it up a little bit before turning their voice into a flute using Tone Transfer and throwing that in the mix somehow. Maybe after all's said and done they could take the whole track and throw it in Audacity to hit Change Speed by -6% a few times to throw some Vaporwave on the thing. No matter how good it ends up sounding I'm sure that's something that would have been otherwise impossible before AI got this good. Plenty of flavors of Plunderphonic Rock music are definitely on the horizon.

As for the fully synthesized tracks like OpenAI Jukebox I'm sure that in a Hip-Hop sampled context they'd work pretty well. If someone wanted to make a Hyper Pop track with built around the hook of a vocal isolated Frank Sinatra synthesized to sing the line "Inject the Mary Jane into my vein!" they could just set the parameters necessary and cook a few versions until one comes out clean enough then feed it through a stem splitter. After that then you've got yourself an admittedly legally dubious sample to throw in a mix somewhere with a little bit of elbow grease. This stuff could quickly result in some really decisive court cases some time from now that may either stifle progress for a while or make a small niche interglobally famous. If things go a little bit awry we might end up with legal precedents for filing copyright on one's general 'vibe' that likely will choke the creative industry even more into disparity between levels of innovation in indie scenes compared to where the money is. Obviously if any big name artist takes note of a track and tries to snuff it there's a much more likely chance it'll Streissand Effect and the person who made the track may just take it as a badge of honor for how well they pulled it off. Either way in the next ten or twenty so years I'm sure there's going to be a fully synthesized artist sneaking onto the scene either from an anarchist lurking in the shadows or from a cynical corporation that may just trick the public for at least a few weeks. It'll be a while before the fidelity is workable enough to pull that off but I can see some lo-fi communities getting fooled sometime from now on the shorter term. Maybe you're reading an A.I. right now?