Sounds can be deceiving these daze. Take the above audio file, for instance. What might sound like fairly respectable and authentic neo-Flamenco guitar music is actually just an earful of AI-generated noodling.
Early this morning I whipped up six Flamenco-centric instrumental iterations at Suno.com. I “created” them all from scratch in about a half-hour. And as per Suno’s written policy regarding tunes generated by Pro or Premier members (costs me $10/month for the Pro plan), I am the legal “owner” of these tracks, and enjoy the benefits of a license for commercial use. Crazy.
Anyone who’s been tapping into my recent posts knows I’m a big fan of Suno.com. Now, Suno is just one player in the burgeoning field of AI music sites, but it’s the original big kahoona on the block, I like how it works and enjoy the results, so I’m trying to suss out how to get the most from the platform before I push on to others… if in fact I ever do.
We face an unavoidable reality: AI can write, produce and deliver a more interesting song than 95% of the songwriters out there (it’s possible I’m being too generous). I’ll be the first to admit Suno can effortlessly write a better, more cleverly-composed tune than I can. AI can no doubt write a better tune than you, and as I see it, we have to either learn to work with AI music and live with it, or else frustrate ourselves past the point of human endurance.
But things are getting a tad sticky in the world of musical AI. Virtually all of the successful AI music sites in existence are being sued for big bucks by various music publishing agencies (eg: BMI, ASCAP, and many others) for inappropriate and/or unauthorized use of copyrighted materials (IOW, infringement).
Their cases center around the contention that AI music sites must necessarily use not just “public domain” material, but also legitimately copyrighted material, for “training purposes” — IOW, teaching the music bots the fundamentals of music as practiced by humans. We’re talking about commonly-used chord progressions and melodic motifs; real-world vocal and instrumental techniques; the various ranges, limitations and common “licks’ used by everyday musicians; the trends and characteristics of countless styles of music — all so that the bots can learn not only how to create sounds humans will (or even can) relate to, but hopefully how to toss together a catchy tune or two in twenty seconds or less!
To be sure, the world at large was not ready for AI! In the realm of music, long-established copyright law was caught completely off-guard, flat-footed, unprepared for — indeed, dumfounded by — what has seemingly emerged in just a few months.
Is it any wonder so many humans are begging for this AI thing to slow down, if not stop entirely. As old-fashioned as I am, I still love my technology… but I’m not sure I am not among them.
Pretty much all experienced musicians can “hear” the commonly used chord progressions and melodic techniques used by virtually every well-versed musician in the creation of their music. And it is just as easy to hear how AI is stealing our licks, so to speak, for the pecuniary purposes of the big-tech savants who design these sites, build and refine them to work so well, not only for musicians but lay folk who couldn’t carry a tune in a bucket.
We’ll be revisiting this topic in the future. Assuming we have one.
