Glenn McDonald “You Have Not Yet Heard Your Favorite Song”: Interview

Glenn McDonald is an interesting guy. He spent the last decade as Spotify’s “Data Alchemist” to shape how music is curated and recommended. I first learned about McDonald through his website “Every Noise at Once”, which is a site that catalogs the entire world’s music. I’ve spent countless hours hyper-fixated on the site listening to jackin’ house, gospel amapiano, Swedish drill, and so much more. His impact on music discovery in the streaming age is immeasurable. 

McDonald’s new book “You Have Not Yet Heard Your Favorite Song” tries to explain the depths, complexities, and conspiracies of the modern music streaming structure. We chatted about streaming, Turkish hip hop, and how K-pop fans ruined data integrity (not really). 

I grew up as a music fan in the LP era as a kid with not very much money. The amount of stuff that I could discover was pretty limited. If it wasn’t played on the radio in Dallas where I grew up, I didn’t know about it. If it wasn’t in the record store, I couldn’t buy it. I graduated from college and got computer jobs and worked on designing software for various things, always dealing with data in some way, like helping people organize it or model it, or something. Never for music purposes, business wise, but I would always use these tools on my record collection database and my music obsession, just because that’s what I cared about. 

At some point in my career, one of my employers got acquired, and I needed a new job. There was this startup [The Echo Nest] a town over from me that some MIT Media Lab people had started that was trying stuff with music data. This was the dawn of streaming, and suddenly, people had a lot more music data than just my record collection. I’d done a bunch of music data projects on my own, so I showed up and was like, “I can sort of do this, could I come work here?” And they were like, “Okay, alright, that sounds plausible.” This startup was called the Echo Nest, and we were trying to do recommendations and music intelligence things for these new streaming services, one of which was Spotify, and they eventually just acquired us, so that we couldn’t do our things for their competitors either. My weird job at Spotify was to try to understand what we could do with all this listening data. And the answer is, we can do a lot with all this.

Along the way I had all these stories I would tell about funny things that happened. Reasons why things were important, or what I found when I investigated problems, questions, fears, or whatever. I’d also read articles and books and everything that people said about streaming and technology. It was always interesting to me, but [I galled] that most of those things were written, inevitably, by journalists. Good journalists, who aren’t involved in doing this stuff, and so they would report on what people said, or feared, or opined about what was going on, but I knew what was going on. I just kept feeling like “Oh, God! Somebody’s gotta write a book about this who A. knows how it really works- not the secrets, but just the mechanics, and B. believes in it; believes in the actual positive potential for this.” Nobody kept doing it. And eventually I was like “Ah crap! It’s me.”

It totally changes your access to the rest of the world’s music. I’ve become totally obsessed with Maskandi, a Zulu folk pop genre from South Africa. There’s no way I would ever have found any of these things. But not only that; because I was into Maskandi, my Maskandi feeds would 

occasionally have these other weird songs that sounded different. I have no cultural connection to this music, and yet it’s amazing to me. There’s no way I would ever have encountered it and if I had somehow encountered one song of it in the pre streaming world, what would I have done with that? I would have had no idea what it was and what context it came from. I would have had no way to find more of it. That to me is giant. It’s on the same level as the Internet itself for listening and appreciating music.

On the artist side it used to be that you basically couldn’t reach people unless you were on a major label. There were very occasional exceptions, but if you wanted to reach me as a listener in Dallas in 1984, you either had to be playing near me at an all ages show, or you had to be on a major label, because that’s all that was played on the radio. I knew about Foreigner, Journey, and Steve Miller, and that was my world. What else was I gonna fall in love with? I didn’t have anything else. 

Now, that technical barrier is gone. I make music. I’m not trying to make a career out of it, but it’s there on Spotify. It’s super powerful that I can get it there and point people to it. I don’t have to press a CD and give it to people. That music is there in the same venue as Taylor Swifts and Ed Sheerans. [Before,] I could dream about it being huge, but it was hard to imagine how that would happen. Now, tomorrow, a million people could click on my song somehow. There’s nothing stopping them. That’s huge, too. You see it in the giant, several orders of magnitude explosion of how much music is at least theoretically available. 

Both of those things are giant. As I say multiple times in the book, I think we’re still very much at the beginning of figuring out how to take advantage of both sides of that, so that our collective experience of music for everybody is much richer. It’s richer for me as a listener, because I spend tons of time on it, and I will go find things. I’m not waiting for Discover Weekly to decide that I might like some weird wedding music from South Africa. I will go find it, I will go wander the globe. But not everybody is willing to make that much of an effort. The tools Spotify have included haven’t done a whole lot to engender that kind of openness and curiosity. They’re very good at giving you more of what you already know. But helping people know- I mean, that’s the title of my book, you may not have heard your favorite song. You have been exposed to what you’ve been exposed to, and maybe those are all things that you sort of love compared to the things that you’ve never heard yet. 

There’s a certain amount of that. But I think ultimately, for me, music and food are the two most powerful things that overcome the innate fear of the other. You imagine people in some place you’ve never been, and it’s easy to imagine that they think completely differently, and their customs are all strange, and their clothes look weird, and you just imagine that their brain somehow works differently and you could never relate to them. And then you taste their food, and you’re like, “Oh, wait! This is really good, and it reminds me of this other thing that I grew up with.” and you hear their music, and you’re like “Oh, they have hip hop, too, and it’s weirdly different, and yet it’s sort of the same.” I think that makes it really visceral, that “Oh, they’re people, too, they’re basically people like me and the ways in which they’re different are interesting more than they are scary.” 

I feel like the next frontier ought to be community. We basically mastered the art of catering to individual tastes. We can feed you more of what you like pretty effectively. It breaks down if you like multiple things, but that could be solved with a little more control on the user’s end. But music is not an isolated thing; music works in communities. Part of the reason that music is great as it brings us together. There’s only little bits of that in the streaming experience right now. Communities can build up around individual playlists and around genres, and at this point, those communities form somewhere other than the streaming service. 

I think that, to me, points to the potential. I want to see communities take shape more clearly on streaming services, on Spotify. I think that’s the way they need to build. As long as the streaming service is maintaining a tight control of what’s happening, and just parceling out listeners to artists and parceling out songs to listeners then it’s hard to make. But if you let communities start to form with more tools, as opposed to just reluctantly allowing them to sneak in via features that you don’t exercise as much control over, then I think it’s huge. 

It used to be music critics. That role used to be DJ’s, and then the role of DJ’s at commercial stations disappeared and it was all programmed centrally. Individual playlist curators, whether they lead with their playlist, or, as in this case, lead with a social presence and make playlists, are doing the roles that reading music reviews in the Village Voice in 1990 did. It’s a whole lot more useful to give someone a playlist than to read a bunch of stuff about an album, and then try to imagine how that person’s listening translated into their prose, translated into your reading, translated back into how this is gonna sound when you listen to it. 

I wrote a music column for ten years. I spent a lot of hours of my life writing music reviews, and there are reasons to do it other than an awkward way of telling people to listen to records. But we don’t need that in order to experience music anymore. The word “influencers” is offensive to me, just ‘cause I’m old, but I think that’s the right way. Getting recommendations from individuals or a community of people who are interested. There’s this metal discussion thread on ilovemusic that I’ve been following for about 20 years. The software hasn’t changed and I check it every day. It has people that I’ve been there with for ages and they’ll alert me to new albums and we’ll discuss whether this band’s new move from death metal to black metal is bad or genius. That’s a little community I get a lot of joy from. 

Working on charts and data integrity, I initially found it offensive, too. I realized, ultimately, the only thing that really bothered me about it was if it screwed up categorization or recommendations. If it interfered with my ability to reflect people’s listening to each other and resulted in reggae fans getting K-pop just because there were too many streams, that was bad. But that was ultimately my problem, not theirs. I had to figure out how to make the math work so that didn’t happen. 

It’s pretty straightforward; they’re sort of cheating? But not really. K-pop fans are not nearly as much of a problem as artificial bots for hire. The K-pop fans are just listening to their K-pop. They’re not listening to a bunch of other random stuff. They’re not in the way, they’re just doing their thing. That’s fine. It’s not really that different from Beatlemania, so grumpy old people being like “ugh, these K-pop kids-” you know what, you were the same way when you were their age about your things. Don’t be judgemental. 

What I said about streaming, that we’re barely at the beginning of understanding what we can really do with all this, that’s ten times as true for AI. You can tell the amount I care about it, because I just took a new job that is AI that’s just “what do we do with this stuff?” In the past two years, our capabilities have exploded, but what’s this good for?

My analogy is: we’ve basically made a museum of all of human knowledge but we’re standing on the sidewalk outside of it, asking the security guard to tell us what’s inside of it. That can’t be the best, most powerful use of large language models, but they’re in this weird state where they’re capable of amazing magic, yet, astounding idiocy. Yesterday, I was trying to figure out how to get AI to answer a question and nothing would work. It would just keep getting it wrong. I’d ask it to explain why it gave the answer and the explanations were stupid. At one point, I said “do your explanations as a limerick,” and it made awesome limericks. The answers were still wrong, its justifications were still stupid, but it made them into limericks. This is both fascinating and maddening. 

In music, I think there’s two places where it’s encroaching initially. One is in AI text-to-music generation and that, to me, is just not interesting. Music is a thing humans do to reach each other. I do not need AI generated music anymore than I need a robot that could eat my dinner for me. It’s not solving a problem. 

What’s gonna happen is that it’s going to become different kinds of instruments. Ultimately, it will be like the move from pianos to synthesizers. This will be the next step from synthesizers to something new, which will be expressive. The transformation is, essentially, type and get a song, but that’s not making music. But an instrument that can turn my creativity and my impulses into music in more flexible and responsive, weird ways, I’m sure will be transformational. As you said, it would just be a tool.

The other part is using AI to generate playlists or recommendations, and so far, that’s pretty underwhelming. Machine learning and AI is most powerful when there are way too many variables for humans to deal with. Drug discovery, gene sequencing, you can’t do that with an abacus. You need computers to do that. Making playlists on the other hand? You’re gonna make a playlist for something and they’re going to listen to five songs on it, they’re going to make a decision every three minutes every 15 minutes. That is not a lot of information. You do not need inscrutable, expensive, complicated processes to do that math and SQL queries. Part of the task is to use this tool for what it’s actually both good for and needed for. 

I still have a lot of ways that I hope the music industry will evolve and I think it is pretty hard to do that within established power structures. Trying to improve the future via an internship at Sony or ten years working at Spotify is pretty hard. Those are big ships and you can nudge them a little bit, and that’s good, people should do that. But I don’t expect any radical changes to the future to come from our giant current giant corporations, because they have massive amounts invested in things being as they are. 

It’s been interesting that in the six months after leaving Spotify, I’ve talked to dozens if not hundreds of startups that have picked some broken piece of the current music industry and are trying to figure out how this really should be done. How can we do this ethically and for the artists’ benefit in a less awkward way, with more visibility or more data, transparency, accountability, or more money. How can we do something better?

Most of those will fail. Most startups fail, most of those artists will crash against the large ships trundling in the shipping lanes and be destroyed. But some of them won’t. So it’s been super interesting to get to talk to a hundred different projects that are very different. I talked to a lot of people at Spotify and my weird role became “the guy you ask when you don’t know who to ask about things.” I got to help with a lot of projects, but they were all within the constraints of Spotify. Now I get to talk to people who have no constraints, like “we’re going to do something totally different.” 

That’s awesome. Some of those things will change the world. Our world will become very different because some little thing will be successful and everybody else will scramble to account for it. 

Optimistic apprehension, or apprehensive optimism. That’s basically the tension I’m going for in the book.