MBW Views is a sequence of unique op/eds from eminent music business folks… with one thing to say.
The next op/ed comes from tech entrepreneur and music veteran Shara Senderoff (pictured inset), Co-Founder & CEO of Jen.
Jen claims to be an “ethically-trained generative AI music platform” that introduces “a brand new normal for copyright compliance in text-to-music era”.
Senderoff is a widely-respected expertise entrepreneur. She beforehand co-founded blockchain enterprise fund and studio, Raised In House, in 2018 alongside Scooter Braun.
Along with Senderoff, Jen named Mike Caren, founder and CEO of Artist Companion Group, as a founding companion final yr.
Jen’s founding staff is rounded out by 20-year expertise vet, Aaron McDonald, who co-founded Futureverse alongside Senderoff and AI PhD Dr. Alex Wang, who led the structure and improvement of Jen’s AI.
Being raised on blues and rock n’ roll provides you a deep appreciation for the human parts of music. The tales. The guts. The soul. “A pc may by no means evoke that,” I wrote in response to numerous inbound decks that I obtained whereas operating my music/tech funding fund that I had co-founded with Scooter Braun. One factor I used to be adamantly not fascinated by was AI-generated music.
This newest AI tech revolution has despatched shockwaves by the enterprise, as a result of we’ve but to see expertise pose a menace to copyrights, this considerably, since file-sharing exploded onto the scene with Napster over 20 years in the past.
It’s essential to acknowledge although, that the actual menace will not be within the expertise itself. It’s within the leveraging of unlicensed copyrights to coach the expertise. All of us had hoped that the times of stealing music had been lengthy gone. However, the rise of generative AI music has ushered in a complete new gang of dangerous actors. The “express regret” later crowd. The “fair-use” military. The “spinoff darlings.” Okay, I simply made that one up. However, it’s correct.
Two years in the past, my staff at Futureverse, a number one AI and web3 infrastructure firm, proposed that we begin to construct Jen. I’ve all the time believed that the sensation of connection that music elicits, each aware and unconscious, is one thing we’ll by no means be capable of comprehend, not to mention replicate artificially. Music is a non secular expertise. How may AI ever get shut?
Our staff of esteemed AI PhDs deeply understood my hesitation. As musicians themselves, they recognized the nuances in composition that generative audio had didn’t seize.
They handed me their first analysis paper (*of many) introducing JEN-1, a complicated text-to-music mannequin designed to beat the constraints of sound high quality in earlier generative music methods. The paper outlined developments that addressed my issues on whether or not or not generative fashions may attain the element of a human composition; an method that captures the nuances of music throughout the total frequency spectrum.
In contrast to AI fashions that depend on visible representations of sound and due to this fact expertise constancy losses throughout audio conversion, Jen’s diffusion mannequin structure generates high-fidelity 48kHz stereo audio, making certain the total richness and depth of sound we’re used to listening to on the radio by working instantly with uncooked audio waveforms. Mixed with multi-task coaching on text-to-music, inpainting, and continuation, Jen permits superior audio high quality and enhanced versatility in AI music creation.
I got here to confidently imagine we may do it technically, however I insisted: we’d do it ethically, or we wouldn’t do it in any respect.
At the moment we launch the alpha model of Jen, a high-fidelity text-to-music generative audio platform that does one thing few have tried but: it doesn’t steal music. With an ethical line drawn within the sand, we architected a coaching doctrine and licensing framework that was unabashedly dedicated to transparency, compensation and copyright identification. Everybody mentioned it couldn’t be carried out, that the music business was too exhausting to work with or that it will take too lengthy and I’d by no means stand an opportunity if I didn’t enter the market first.
“All of us had hoped that the times of stealing music had been lengthy gone. However, the rise of generative AI music has ushered in a complete new gang of dangerous actors.”
Jen has over 40 fully-licensed catalogs in its preliminary coaching set. It’s ethically-trained and introduces a brand new normal for copyright compliance. Our rigorous coaching course of and diligently architected method to doing it proper is unmatched. Each monitor is mechanically vetted for audio recognition and copyright identification using a database of 150M tracks. This contains each the compositions within the coaching set and each newly generated monitor on the platform. We couldn’t do that if we stole music.
Moreover, Jen generates a cryptographic hash for every monitor that’s then recorded on The Root Community blockchain. This course of supplies a sophisticated type of verification, making certain the integrity and timestamp of every monitor’s creation. Verified music outputs obtain a JENUINE™ indicator upon era.
Right here’s what this implies and I’m going to do the unthinkable… I’m going to reference Taylor Swift. If you happen to kind in a immediate on Jen that claims “make me a music that feels like Taylor Swift,” it’s not doable in any respect for it to do something of the kind. It has by no means heard Taylor. So how may it know what she feels like? If a nasty actor provides a Taylor music pattern or vocals decoding her melodies or lyrics to a Jen monitor after its creation, upon overview, our hash will point out that the monitor was altered post-generation and ensure Jen’s output didn’t comprise the pattern. Once more, we will do that as a result of: we don’t steal music.
Shoppers outright personal the tracks they create on Jen. They will exploit and distribute them throughout DSPs, socials, and so forth. You realize why? As a result of we don’t steal music.
The business response to generative music has been to threaten takedowns for something created utilizing AI. However generative music expertise will advance at speeds far larger than any expertise able to patrolling it. This answer is basically flawed. The business will likely be mortified when they’re caught distributing the tracks they’re combating towards with out even figuring out it. As an alternative of making an attempt to cease it, what if we as a substitute labored collectively to steer its course?
We have to defend the worth of music and the respect it deserves all over the world. We’d like sturdy licensing frameworks that safeguard copyrights and enact penalties for firms that use unlicensed music for coaching information. We’d like the business to talk up now. Not tomorrow. Not after forming an AI committee. Not subsequent yr. Now.
Many concern that AI expertise poses a menace to artistry. However the actuality is, expertise has impacted music for hundreds of years and we haven’t but misplaced artistry. It’s not going to switch artists or the artistry we admire. It’ll collaborate with them. Empower. Improve their work in new methods. Taylor would nonetheless be Taylor if she was arising now. The magic and hearts of musicians will solely be amplified by the tech arising. So long as we carry the hammer down.
I don’t suppose there’s been a second that might outline the way forward for the music business extra so than this one. Irrespective of how troublesome it’s to do that the appropriate method. Somebody has to show it’s doable.
My hat’s within the ring and I’ll present you that it may be carried out with out stealing music.Music Enterprise Worldwide
