MBW Views is a series of exclusive op/eds from eminent music industry people… with something to say.
The following 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 new standard for copyright compliance in text-to-music generation”.
Senderoff is a widely-respected technology entrepreneur. She previously co-founded blockchain venture fund and studio, Raised In Space, in 2018 alongside Scooter Braun.
In addition to Senderoff, Jen named Mike Caren, founder and CEO of Artist Partner Group, as a founding partner last year.
Jen’s founding team is rounded out by 20-year technology vet, Aaron McDonald, who co-founded Futureverse alongside Senderoff and AI PhD Dr. Alex Wang, who led the architecture and development of Jen’s AI.
Being raised on blues and rock n’ roll gives you a deep appreciation for the human components of music. The stories. The heart. The soul. “A computer could never evoke that,” I wrote in response to countless inbound decks that I received while running my music/tech investment fund that I had co-founded with Scooter Braun. One thing I was adamantly not interested in was AI-generated music.
This latest AI tech revolution has sent shockwaves through the business, because we’ve yet to see technology pose a threat to copyrights, this significantly, since file-sharing exploded onto the scene with Napster over 20 years ago.
It’s important to recognize though, that the real threat is not in the technology itself. It’s in the leveraging of unlicensed copyrights to train the technology. We all had hoped that the days of stealing music were long gone. But, the rise of generative AI music has ushered in a whole new gang of bad actors. The “ask for forgiveness” later crowd. The “fair-use” army. The “derivative darlings.” Okay, I just made that one up. But, it’s accurate.
Two years ago, my team at Futureverse, a leading AI and web3 infrastructure company, proposed that we start to build Jen. I have always believed that the feeling of connection that music elicits, both conscious and subconscious, is something we’ll never be able to comprehend, let alone replicate artificially. Music is a spiritual experience. How could AI ever get close?
Our team of esteemed AI PhDs deeply understood my hesitation. As musicians themselves, they identified the nuances in composition that generative audio had failed to capture.
They handed me their first research paper (*of many) introducing JEN-1, a sophisticated text-to-music model designed to overcome the limitations of sound quality in previous generative music systems. The paper outlined advancements that addressed my concerns on whether or not generative models could reach the detail of a human composition; an approach that captures the nuances of music across the full frequency spectrum.
Unlike AI models that rely on visual representations of sound and therefore experience fidelity losses during audio conversion, Jen’s diffusion model architecture generates high-fidelity 48kHz stereo audio, ensuring the full richness and depth of sound we’re used to hearing on the radio by working directly with raw audio waveforms. Combined with multi-task training on text-to-music, inpainting, and continuation, Jen enables superior audio quality and enhanced versatility in AI music creation.
I came to confidently believe we could do it technically, but I insisted: we’d do it ethically, or we wouldn’t do it at all.
Today we launch the alpha version of Jen, a high-fidelity text-to-music generative audio platform that does something few have attempted yet: it doesn’t steal music. With a moral line drawn in the sand, we architected a training doctrine and licensing framework that was unabashedly committed to transparency, compensation and copyright identification. Everyone said it couldn’t be done, that the music industry was too hard to work with or that it would take too long and I’d never stand a chance if I didn’t enter the market first.
“We all had hoped that the days of stealing music were long gone. But, the rise of generative AI music has ushered in a whole new gang of bad actors.”
Jen has over 40 fully-licensed catalogs in its initial training set. It’s ethically-trained and introduces a new standard for copyright compliance. Our rigorous training process and diligently architected approach to doing it right is unmatched. Every track is automatically vetted for audio recognition and copyright identification utilizing a database of 150M tracks. This includes both the compositions in the training set and every newly generated track on the platform. We couldn’t do this if we stole music.
Additionally, Jen generates a cryptographic hash for each track that is then recorded on The Root Network blockchain. This process provides an advanced form of verification, ensuring the integrity and timestamp of each track’s creation. Verified song outputs receive a JENUINE™ indicator upon generation.
Here’s what this means and I’m going to do the unthinkable… I’m going to reference Taylor Swift. If you type in a prompt on Jen that says “make me a song that sounds like Taylor Swift,” it’s not possible whatsoever for it to do anything of the sort. It has never heard Taylor. So how could it know what she sounds like? If a bad actor adds a Taylor song sample or vocals interpreting her melodies or lyrics to a Jen track after its creation, upon review, our hash will indicate that the track was altered post-generation and confirm Jen’s output did not contain the sample. Again, we can do this because: we don’t steal music.
Consumers outright own the tracks they create on Jen. They can exploit and distribute them across DSPs, socials, etc. You know why? Because we don’t steal music.
The industry response to generative music has been to threaten takedowns for anything created using AI. But generative music technology will advance at speeds far greater than any technology capable of patrolling it. This solution is fundamentally flawed. The industry will be mortified when they are caught distributing the tracks they are fighting against without even knowing it. Instead of trying to stop it, what if we instead worked together to steer its course?
We need to protect the value of music and the respect it deserves around the world. We need robust licensing frameworks that safeguard copyrights and enact penalties for companies that use unlicensed music for training data. We need the industry to speak up now. Not tomorrow. Not after forming an AI committee. Not next year. Now.
Many fear that AI technology poses a threat to artistry. But the reality is, technology has impacted music for centuries and we haven’t yet lost artistry. It’s not going to replace artists or the artistry we admire. It’ll collaborate with them. Empower. Enhance their work in new ways. Taylor would still be Taylor if she was coming up now. The magic and hearts of musicians will only be amplified by the tech coming up. As long as we bring the hammer down.
I don’t think there’s been a moment that could define the future of the music industry more so than this one. No matter how difficult it is to do this the right way. Someone has to prove it’s possible.
My hat’s in the ring and I’ll show you that it can be done without stealing music.Music Business Worldwide
