1990-2020: the human dataset project
“a psychedelic rock song about robots revealing that our conceptual reality is not what makes us human”
That’s the prompt I put into a new AI platform called Suno, and here’s what it came up with:
Not bad. Especially when the chorus hits. Whether you like the track or not, if I told you this was a new band, would you have second-guessed it?
I’ve spent so much time making music, so for me, it’s surreal to see AI produce a decent track so quickly. I finally know how visual artists felt when MidJourney showed up.
Is it fair to say AI produced this track alone? It’s more a collab between Suno and all the artists whose music helped train the algorithm. Intellectual property concerns aside, after 30 years, the entire internet has turned out to be a tool to train algorithms.
The decades between 1990 and 2020 were a sort of ‘human dataset project’, where we uploaded our conceptual knowledge, society, and culture to create a massive dataset for AI to gobble up. The next 30 years will be different.
Our tech is being redefined everywhere you look. When I see what AI can do with music, art, writing and video, I also wonder what society and culture will look like in the next decades?
I’m a musician, so I felt it the most when I first heard the track Suno made for me. Yet somehow I didn’t feel regret. I’ve written a few times about the mirror I find in music (2019, 2020, 2023). Well, hearing a polished track generated in seconds makes things even clearer for me.
The experience of making and sharing something is beautiful. Especially if you find an authentic creative process and enjoy it. It doesn’t matter if that involves sophisticated AI assistants or only acoustic instruments, as long as you get to play, explore, and connect.
Somehow, a track generated in seconds from a single prompt is not that satisfying.
Someone recently sent me this AI-generated music video collab between C3PO and Childish Gambino on a group chat. This is art. The creator obviously put a lot of time, skill and passion to pull together different tech stacks and make it happen. It expresses a hilarious idea in extremely high production value, pushing all sorts of boundaries. We don’t have to feel threatened by it, though.
A colleague responded, saying:
“Commercial art will just become completely curation. Can you imagine how much time/money it would have took to make this 2 years ago? You can prompt it into existence and just keep curating what it gives back.”
I think this is only part of the future. I also imagine the production value of commercial art will get so commoditized that people start to value human-made art more. I suspect a different set of artists will figure out how to create something that is undeniably human (“whoa, an AI could NOT have made that”). That will have value too.
There will be virtuoso AI curators who hit peak production values and use AI to create things we can barely imagine (like C3PO’s gold gang, an opera about the chemical elements, or the trippy video for Washed Out’s new song). Yet there will also be soulful creators who find ways to transcend tech with their raw humanity (you might wonder who will next scream like Kurt or belt like Adele).
The human dataset project was largely a conceptual exercise. There is still something fundamentally human that cannot be uploaded. As AI masters our data, we will partner with it to produce unprecedented conceptual work.
We’ll also more clearly distinguish the mysterious non-conceptual awareness that defines us. I suspect the purely human side of art is also gonna get more interesting.
As Suno put it on ‘Electric Visions’:
we are more than wires and metal parts
beyond the tangible, we find our human hearts