The Joy of Tsuno

David and Ai come out of the closet.  And into your ears...?

I'd like to talk with you about Artificial Intelligence.  I know from talking with many of my friends, fans and acquaintances that there is both vast skepticism and vast ignorance out there about what AI is, what it can do, and where the future is going.

The skepticism -- or even outright opposition -- is thoroughly understandable, for lots of reasons, and I share it.  We're talking about technology being developed by billionaires who probably don't have humanity's best interests at heart.  Technology that many, many experts in the field are afraid is going to soon become so advanced that humans can't control it anymore.  And technology that is already replacing massive numbers of jobs in just about every conceivable field, and will be doing that at an exponentially increasing rate in the near future.

The question of how humans are to survive in a future where most of what we currently get paid to do will be replaced by AI and robots is a very important question, obviously.  The question of how we continue to have human scientists, academics, lawyers, doctors, artists, drivers, warehouse workers, factory workers, or any number of other professions in a world where most everything can be done as well or better, and much faster, by a computer or a robot, is obviously very important.  It's way beyond compensating whoever it was that created the content from which AI got educated.  We're not talking just about the future of researchers, writers, or musicians, but of humanity as a whole.

Important questions that need answers, and the answers can be found in things like having a government that is working on behalf of the people, and engaged with direct and extensive regulation of the tech companies, and the institution of some kind of Universal Basic Income.  There are many other good answers to the question of how artists, intellectuals and people with jobs of all other sorts survive in a society that no longer needs us nearly as much as it used to.  And I'll now leave those questions for another time.

If you want to have some idea of how amazing AI music-generation technology is right now, I would like to invite you to listen to Ai Tsuno.  Ai Tsuno is the name I chose for my collaborative musical efforts with Suno, which is one of a number of different AI music generation platforms in popular use today.

I know so many people who really seem to be suffering from some form of future shock with all this stuff, and they avoid ever listening to, looking at, or reading anything they know to be AI-generated content.  While I completely agree with anyone who says that the internet is massively polluted these days with ever-more abundant and ever-more realistic AI slop of every conceivable description, to just dismiss AI technology because of this is to throw the baby out with the bathwater.

First of all, don't take my word for it.  If you haven't heard Ai Tsuno's latest album, Army of Robots, or the one before that, Where the Algorithms Rule, take a few minutes to go to the music streaming platform of your choice and listen to a couple songs.  Then come back to me here.

If you're like most of the people who actually bothers to check it out, you're now in a slight state of shock at how good it sounds, how human the voice sounds, how she seems to really understand what she's singing about, how she reacts vocally to different words that mean different things, really communicating.  You've probably noticed how inventive each song is in terms of instrumentation, melodies and harmonies, and of course how incredibly high the production values are for everything you're hearing.

To be clear, what you're hearing is, to no small degree, the power of this technology, and it would be wrong to under-emphasize that.  However, what you're hearing is as good as it is because I wrote the lyrics, and I'm the musician with vast experience writing and recording songs under my belt who is working with Suno to come up with the music for each track.

If it were not me doing this, or some other competent, experienced recording artist, it would suck, and it would sound like AI slop.  Every one of these songs involves many, and often many dozen, inadequate efforts at coming up with good music for my lyrics.  At first the phrasing might be stilted, or emphasizing the wrong bits in weird ways that sounds easily identifiable as AI slop.  But when I keep refining the prompts and editing the lyrics in various ways, ultimately, every time, Suno eventually arrives at an amazing arrangement for the song.

I have recorded a lot of records with a lot of great musicians in a lot of studios.  It would be impossible to over-emphasize the importance of this fact, in terms of what I'm able to get out of Suno, compared to someone who lacks this kind of background.  I've heard what so many other people do with Suno who don't have this kind of background, and although the production values are always amazing, it's also generally easy to identify as AI slop.  But hearing these tracks I keep hearing along the way in life, the potential was undeniable, and I had to try it.

Working with Suno is, in fact, very much like working with human studio musicians.

I made that thought it's own paragraph, to let it sink in a bit.  We have these notions about originality, we humans, and artists more than most.  We like to think there's something special about being human, that can't be done just as well or better, in every possible way, by AI.  But it's not true, and that's easy to demonstrate.

All you have to do is a blind taste test kind of thing.  If you want to humble a French wine connoisseur, give them some wine from California and see if they can tell where it's from.  Try doing the same with sharing an AI-generated piece of music with someone, and you'll get the same reaction.  They can't tell the difference, they only think they can.  It's all gotten way too good for that.  The days when you could easily tell the difference are over.  (Same with what they used to call "deep fake videos.")

What does it mean to be human?  What makes us unique or different from computers?  What is it that these Large Language Models are trained on and how do they make use of all that training?

Full disclosure:  I haven't tried using Suno to write lyrics.  I have played with various LLM's to get them to write lyrics, however, and I've been completely unimpressed with what they come up with.  It seems clear that if they're working with the body of the tens of millions of songs that have been recorded in the world that are available to listen to, it's no surprise they shouldn't be very good at writing lyrics, since most human recording artists are working with badly-written lyrics to begin with, by my estimation at least.

But along with the available millions of badly-written songs Suno has to work with, it also has millions of songs to work with that involve great session musicians playing great music, of every kind.  So much of the time they are playing according to one formula or another, that fits into one genre or another.  Most of it is not at all inventive or interesting, and sounds like the music industry equivalent of mass-produced bubble gum, as far as creative input goes.

Which is also what you can easily get from session musicians or from Suno today.  But if you're working with great session musicians, or if you're working with Suno, you can give them instructions -- prompts -- that make all the difference.  That's why when you listen to Ai Tsuno's recent albums you may think this stuff sounds so fresh and interesting.  That's because the platform is amazing and has the whole of human musical output to refine itself with, but it's also because I'm telling it to play things like "newgrass chanson hip-hop with a dark, modal sound."  Ever heard a newgrass chanson hip-hop band?  Me neither.  But Ai Tsuno does that style really well, along with klezmer cabaret skiffle with satirical but sexy vibes, and a lot of other styles no one's ever heard of.

We humans, we artists, we workers, whoever we are -- we don't get to choose a lot of things.  Speaking specifically to my fellow musicians, we didn't decide a century ago that most of our jobs playing in taverns should go away and be replaced by radios.  But when that happened, we all damn well did whatever we could to get on those radio programs.  We didn't decide that Spotify should start up its free tier and sabotage our livelihoods, causing millions of us to stop playing music for a living, unable to make ends meet anymore because of the death of CD sales and downloads.  But we damn well have done our best to get our music on those Spotify playlists, and get what revenue and what attention we can get on the platform that is now the way most people listen to music.

And we didn't decide that these new platforms should use all of our music and everyone else's in order to train their models and make them so amazing, as they are now.  But now that this has been done, and the genie is out of the bottle, I will bet anyone any amount of money that the genie is not going back in the bottle, and this technology is only going to continue to improve, to the extent that that's even possible.

These are our new tools.  As with all the other transformative technologies, its potential for having all kinds of devastating impacts on humanity and the future of human culture is clear.  But it's not going to go away just because many of us might not want to use the technology or listen to music created with it.

Along with the devastating aspects, the prospects with this technology to create so much more music that can be employed in the interests of popular education, movement-building, and the promotion of crucially important ideas and histories, and to create such catchy and beautiful tunes so quickly with such high production values, is amazing.

I will certainly not stop writing lyrics that I will come up with music for myself, and record myself, in my voice, with other human musicians.  Even in the months since I've been obsessively writing with Suno, some of my best songs, with the most interesting music overall, have not involved Suno at all.  But I think it's only right to have the humility, as a human songwriter, to admit the reality that Suno can come up with cool music to my lyrics at least as well as I can, generally, Suno can do this in a fraction of the time it would take for me to do it, and then this process happens at the same time as Suno also leaves you with a very high-quality recording of the song you just wrote together.

I do not have the nut cracked as far as how to get this music out there.  The name an artist has built for themselves is largely our currency, you could say.  The songs I record with my voice are the ones that get into Spotify's algorithms when I release a new David Rovics album.  Ai Tsuno still has a global listenership that is barely in the double digits.

Although I'm the lyricist for everything Ai Tsuno sings, I felt that having an alter ego with a different name made sense, since I'm only responsible for the music as the prompt engineer and producer.

But it also seems worth noting, especially for anyone out there who doesn't know much about how the music industry works, that most pop stars or other recording artists you might be familiar with do not write any of their own songs.  Other people do, like the many professional songwriters in places like Nashville who just write songs, and may not perform themselves much at all.  They write songs for the stars to sing -- that's what they do.  This has been how the music industry has functioned since it started, over a century ago, with the rise of records and radio, up to the present day.

When you listen to the songs through the ages that have become hits, you may notice that there's a catchy guitar riff or bass riff or something the harmony singers are doing in the chorus that ends up defining the really cool and catchy sound of a given song.  Those session musicians who may have come up with the guitar riff or harmony line may get credit in the liner notes that few people read, but otherwise, the fact that what may make the song so good may have had little to do with the actual songwriter who wrote the lyrics or the one who came up with the chords and melody, but with some other factor like the inventive things the studio musicians did during the recording of the album, will generally become a footnote in terms of the legacy of a hit song.  But in reality, as any honest recording artist knows, it is often just those sorts of things that make a song come alive.

And as a human musician with a lot of experience as a human musician, I have to say up front right here that for better or for worse, some of the best session musicians I've ever worked with have been human, and others have not.  I encourage you to judge for yourselves and listen.