‘You'll notice I do not call it AI music, because I don't think it's music’: Qobuz’s Managing Director Dan Mackta on music streaming’s biggest problem, and how he hopes to tackle it

Things music-lovers enjoy about Qobuz: it was the original hi-res music streaming service and it also offers 24-bit hi-res downloads on demand (ie., no need to subscribe to buy) alongside a regular streaming model. Then, there's the digital Qobuz Magazine, human-first playlists over algorithms, a more ethical payment at significantly higher average per-stream royalties than mainstream competitors, the fact that it refuses to run a free ad-supported tier that devalues music, it continues to transparently publish its financial payout data, it recently added in-track lyrics… I could go on.

But I'm not here to talk to Managing Editor Dan Mackta about any of those aspects of Qobuz today, admirable though they are. No, I want to talk to him about generative AI, and Qobuz's strong stance on it, as set out in its February-release AI charter.

If you haven't read the document, the following quote may be enough for now. It reads as follows.

Dan Mackta has made time to speak to me from a hotel room in Connecticut, at 9:30am local time, before setting off on the 100-or-so miles to Rhode Island and the Newport Folk Festival.

Anyone who speaks to Mackta will find a music lover from the Big Apple first and foremost — a businessman not afraid to talk openly and honestly about an industry he knows and loves deeply.

Right now, Mackta is hoping the rumors are true and that Lauryn Hill will indeed make a rare appearance to play a set (she did), adding with no small amount of excitement, "I've wanted to go my whole life, and this is the year it's finally happening".

Ready or Not

Yes, all of these sub-heads are now going to be Lauryn Hill tracks, because I'm jealous of Mackta's upcoming weekend of live music. I ask Mackta how long Qobuz has been concerned with AI-generated audio infiltrating the platform's otherwise notably human-centric offering. It's clear he's not a fan.

"It was a couple of years ago that we noticed the volume of AI content, fake music, whatever you want to call it. It was a problem from the beginning because of the volume of stuff that was being delivered.

"There was already a fraud problem in music streaming — basically content uploaded just for the purposes of fraudulent streaming, money laundering, organized crime and all of that crazy stuff — but generative AI allowed scammers to create this stuff at crazy scale. They do it to hide their tracks; create so much stuff that nobody can even really tell what the heck is going on. So we knew it was a problem.

"We also knew that this content did not fit into our vision for what our music service should be focusing on. So early on, the guys in France — because all of our tech is in France — looked at a few detection solutions that were on the market at the time, about two years ago, and decided to create their own detection methodology.

"So, we started analyzing. The thing is that the analysis takes a long time, so it's being rolled out in stages. All new releases are now analyzed."

Mackta is also refreshingly open about where the company's at with the rollout. "The actual tagging piece of it, which is where we notify the user if something has been detected as AI, hasn't rolled out yet, although it's planned for September. We shall sell no wine before its time!"

DDEX (That Thing)

I ask Mackta about the relevance (or effectiveness) of voluntary AI self-identification as a solution to the problem of 100% AI tracks — the 'Hey, here's an AI content badge; we've created it and given it to labels and distributors, so they'll use it' approach, because it's one we've seen in the form of 'Transparency Tags' on rival sites. Mackta is quick to clarify that actually, it is a new development and does have some value.

"I mean, the DDEX, (Digital Data Exchange, which is the metadata standards-setting organization upon which the music industry is run) recently added some fields to its structure. But it's basically voluntary for the person setting up the music for release.

"They fill in these fields in the DDEX. So it's 'you' and 'title' and the ISRC (International Standard Recording Code — a 12-character unique digital fingerprint that identifies a specific sound recording or music video) and UPC (Universal Product Code, which is similar but for the whole product) and other areas for credits. It's a whole elaborate standard.

"So, they added fields for AI where you're supposed to say if something is 100% generated by AI or if the lyrics were AI-generated.

"Now, some of the DSPs (Digital Service Providers, ie., rival sites) said, ‘All right, we’re starting to get this information from our suppliers, so we're going to surface it in a way that the users can see’ — and we'll have that too; we get the same information. But obviously, big problem: why would a scammer admit? If it's voluntary, it's a partial solution."

The Miseducation

Mackta is keen to impress upon me that while detection is a big part of the issue, it's not just about AI detection when it comes to Qobuz — or for the wider music industry. The company has thought deeply about what it's going to do with this content. "That's why the AI charter was a very comprehensive doc; it wasn't just about AI content", Mackta confirms.

What is the bigger picture? "It was about our philosophy as a company about AI and as a tool in our business — our everything. When we're going to use it and when we're not, and transparency around that. So that was about more than just the AI content."

Mackta sighs. He's ready to pin his colors to the mast. "You'll notice I do not call it 'AI music' because I don't think it's music. I think it's a simulacrum of music to get, you know, critical theory. But it's… it ain't music."

Everything is Everything

I want to know more about the different types of machine learning in music we accept, versus generative AI audio for fraudulent use.

Mackta is incredibly open about the versions of machine learning Qobuz does use. "Listen: there are a lot of tools in music creation workflow that have been around for a long time that are essentially versions of AI. But it's very different when you're using creative tools as a musician versus a scammer saying, 'Give me 10,000 things that sound like music that I can upload, for fraud.'

"Our number one most requested feature was autoplay. Right? When your cue ends, the Qobuz user wants the music to keep playing. So, that's machine learning! That's AI that looks at what you were listening to, and runs an algorithm to figure out what it should play next.

"So, we have that. Users wanted that. We have a few playlists now that are algorithmically generated based on your listening and your favorites. But that's like two clearly marked playlists, and if you don't want that, you never have to look at it.

"The thing is that some people — especially as Qobuz has been getting thankfully bigger and starting to poke into the mainstream — those people are used to those kinds of features and they don't realize it’s AI. We'll always make it clear what's been selected by humans versus what hasn't."

Lost Ones

I posit the idea that some have suggested — that AI is here, it's not going away, and there's an argument for not even trying to stem the slop. Mackta laughs, but he gets it. "I mean, let's be real. We're doomed! But let’s also try to hold it at bay, however we can.

"To me, the whole thing is just a really, really big bummer. You know, it just is so different from ‘why am I in the music business’? Why, really though, do I do this? Having to deal with this issue on top of all the other issues that were already there in music just makes me really sad, and so much time and energy and resources wasted.

"It’s largely an unintended consequence of the streaming business model, with the way revenues are shared; it's always been a target for fraudsters. It's like if every man, woman, child, and dog released three albums last year on the planet, it still wouldn't be as much music as the fake music that’s being shoved down our pipes."

Ex-Factor

I muse that perhaps as long as it's human beings making content for the emotional appreciation of other human beings (which is what Deezer's head of research recently offered, when I interviewed him), perhaps we'll be OK? Mackta is quick to jump in. "Well, and that's what music is! It's the whole thing that makes it all go!

"It's unique. It's a form of communication that taps directly into emotions of humans. So we can take humans out on either side of that equation — it ain't music anymore. What makes music music? You know, if you ask those questions, I think it all becomes pretty clear that this AI stuff may sound like music, but…"

Quite right. So what's the future for Qobuz? And can he speak to the music industry at large? Mackta is less certain, but he definitely isn't keen on one possible development.

"It's not really clear what's going to happen with this generative AI stuff. I thought the music industry would do its usual thing and try to stamp it out. But instead, now these companies are basically licensed by the major labels; if everybody's getting rich off of it, they're not going to try to stamp it out.

"But I do think, in terms of the fraud piece (of the puzzle) and the massive amounts of content that's just being generated purely for the purposes of fraud, the industry is going to come to some kind of resolution about it, because it's just not kosher.

"I don't know exactly what it's going to look like, but Qobuz is part of a trade group, the Music Fights Fraud Alliance (along with other DSPs and distributors and aggregators), and we're all trying to come together — the way the banking industry has come together to share certain information to be able to stop fraud. It's kind of like herding cats."

All Falls Down

And there's another 'what if?' Mackta wants to share. "At Qobuz, we sometimes talk about, like, what if AI fake music becomes something that most people want to listen to, you know? What if that's what's considered music in 10 years? Are we still going to have our philosophy? I would say I hope so. I hope that we stick to our guns and the stuff that makes music special is not going to change.

"And I do think that 90-whatever-percent of people probably could be played AI content and be told it's music, and they wouldn't know. They wouldn't care! And that's a shame. But that's reality.

"However, for people who are just even a little bit into music, I don't think we're going in that direction. I think, if anything, there'll be a backlash coming soon that really values human curation, human creation, and actual artistry and emotion and communication. You know, you just can't fake it."

I tell Mackta that I agree, and that as an occasional performer myself I really hope he keeps up the good fight. He laughs. "Yeah, I mean, if anything, it's gonna force us to dig in even more on these principles. And from a business standpoint, I think that's just fine. I think that success for us is going to be in how we differentiate and stay focused on what we think is important. And so far, so good!"

Agreed. As a loyal Qobuz fan myself, I'd go so far as to say it's a case of so far, so very good.

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

The best turntables for all budgets

➡️ Read our full guide to the best turntables
1. Best overall:
Pro-Ject Debut Carbon Evo
2. Best cheap beginner deck:
House of Marley Revolution
3. Best budget Bluetooth deck:
Audio-Technica AT-LP70XBT
4. Best premium Bluetooth deck:
Cambridge Audio Alva TT V2