Why goofy AI art almost never seems wasteful to me

Environmentalism is important, but shouldn't be used to squash pluralism

Good art, bad art, and pluralism

The latest thing I made with AI is especially tasteless and frivolous. I’m a big fan of Sufjan Stevens, and also The Dropkick Murphys, and thought combining them would be fun:

I’ve listened to this six times today.

AI music has been great for people like me who can enjoy more tasteless and goofy stuff. Recent debates where AI art gets called “soulless slop” have felt personal as someone who enjoys soulless slop. I feel uniquely happy sitting in an Ikea-furnished room with lofi beats on.

There’s an interesting ethical question of whether it’s justifiable to use any energy to make bad art, given the climate crisis. There are two extremes in this debate.

One extreme says “Art needs to meet a specific threshold of tastefulness and seriousness and humanness to justify spending energy on making it.”

This sounds authoritarian and conservative. A lot of my favorite media, like pop art, wouldn’t have made the cut if this rule were applied in the past.

Is this tasteful and human and serious?

On the other side is “We should just pursue everything we want to do without any care for how much energy we use. Freedom!”

This is also a very bad rule. Climate change is real and bad and our personal consumption choices matter.

Source

I want a rule that respects both the reality of climate change and deep, deep cultural pluralism. Pluralism is really important to me. I want to live in a society where people have strong cultural approval to enjoy “high art” like Anthony Braxton and Arvo Pärt, and “low art” like Doja Cat’s Moo and Clarence Clarity and AI Plankton singing Pink Pony Club. If the climate costs of each are the same, low, and there are people who like them, we shouldn’t bring quality into the debate about whether art can justify its own existence. That would be anti-pluralist and artistically conservative. It disturbs me when climate change is used as a way of enforcing cultural conservatism.

My rule for when it’s okay to spend energy on making something

I think these two questions are a great way to judge the energy we spend on art in general:

1. Is it okay that this thing exist, or is it bad to spend literally any energy making it?

2. Can we make it using an acceptably low amount of energy?

If more people were consistent in trying to answer these two questions, I think most complaints that AI images and music is wasteful wouldn’t be made.

Here’s how I’d apply these rules for my Dropkick Murphys Sufjan Stevens cover:

1. Is it okay that this thing exist, or is spending literally any amount of energy on it unacceptable?

Imagine AI doesn’t exist. You find out I spent a few hours on GarageBand playing around with sounds, modifying my voice, and downloading musical samples from the internet, and at the end I had what sounds like the Dropkick Murphys covering Sufjan Stevens. I upload it to the internet. It gets a few views, but most of them are from me. I listen to it for a few weeks and then get tired of it and move on.

Would your first thought be “He really shouldn’t have done that, we’re in a climate crisis”?

This would seem like a huge overreaction. It’s okay that we spend some energy on silly frivolous stuff, as long as we’re not using a crazy amount. It’s okay that a lot of goofy and bad music exists on YouTube and Spotify. This is a funny side-hobby people have that isn’t really contributing much to the climate crisis. My Dropkick Murphys Suftan Stevens cover fits this category. It’s okay to spend some energy to make it. The next question is how much.

2. Can we make it using an acceptably low amount of energy?

A back of the envelope calculation says that an AI song probably uses between 1-8 Wh to generate. Let’s assume 8 Wh. Making the image for the cover art probably took another 3 Wh, bringing the total to 11 Wh.

How much energy would it take to make the song on GarageBand?

I think I could make this song with something like 5 hours of work on my laptop. Assuming my laptop is especially efficient and uses 30W, this means that making this song would have used 150 Wh. Over 13 times as much energy as AI.

It’s often very hard to find a situation where making something with AI takes even 20% as much energy as making it most other ways, just because a laptop’s normal energy is (at minimum) 30 Wh per hour, and the energy cost of physical material is even higher.

So, given that it’s okay that we spend some energy to make an image or music, it’s almost always going to be better for the climate to use AI to do it. This isn’t to say that we should only use AI (I mostly don’t consume AI content and don’t expect to). Climate shouldn’t be our only consideration in making stuff. But if you’re concerned about climate, AI is probably never going to be worse for producing any media you want than using a laptop to make it.

My Dropkick Murphys cover passes both tests. It’s fine that I made it, and AI used way less energy than other ways of making things we normally consider totally acceptable (GarageBand etc.). Quality shouldn’t be part of the debate.

Conclusion

A lot of talk about AI images and music being “wasteful” seem somewhat anti-pluralist, and not really numerate in how much energy different methods for producing art use. When I see people criticize goofier or “tasteless” AI art for being wasteful because it’s goofy and tasteless, I feel like I’m talking to an artistic conservative. I don’t want climate change to be used to push anti-pluralism. People should be free to pursue what they like in the way they like it. We should have strong cultural norms that allow this.

There are a lot of good reasons to worry about AI art and music:

• The general quality of art we see every day going down as businesses cut corners.

• Risks to artist and musician jobs.

• Fewer people learning to make art to the point that art makes less progress in the future.

• Fewer first-hand experiences of making art, which is valuable in itself.

But arguments that use climate as a way of squashing people’s pluralistic tastes are bad. I and others who love “low” art deserve the right to consume and produce it without getting condemned, as long as we’re not emitting too much as we do it. AI is basically always going to clear that environmental hurdle compared to every other way of making (in this case, often bad) art.