Wireborn Art Isn't Art

In the past few days I’ve seen two different stories written by GPT-5 shared by smart people who feel these are very good stories. I cannot find it in myself to care enough to even read the first paragraph of these, and it’s not for the typical AI-hate reasons that are so popular right now. It’s because the wireborn¹ can’t make art of the kind that’s worth interacting with.

To explain, let me quote Yglesias’s sickest burn on one of my favorite old bands.

Emotional Participation - Imagery

1990s grunge band Bush gets a lot of flack for nonsensical lyrics, but lyrics aren’t the point of a song. The goal of art is to take an emotion within the artist and create it in the audience. Lyrics are one part of that, and they don’t need to be plainly scrutable to achieve that goal. As long as they reliably push the listener into a state of feeling the target emotion, they are doing their job.

In fact, less direct words that force the listener to bridge the divide between the actual words to the imagery they point to is preferable because it involves more neurons and forces active participation in the art rather than docile reception. “The sky above the port was the color of television, tuned to a dead channel” is a far better line than “The sky was cloudy and grey” because it forces you to see an image of a static-y TV and feel what seeing that is like, rather than just pinging the “cloudy” and “grey” tags in your vocabulary. It is engaging the machinery of your imagination and forcing it to do some work, entangling it with the experience of reading. This shifts the language from the literal (cloudy, grey) to the poetic (television tuned to a dead channel).

To do this there must be enough shared context that the image created by the author reliably evokes the desired emotions in readers. Author and reader must share similar emotional attachments to the same sense-memories, or the writing doesn’t work. The theme of Neuromancer is isolation and alienation, and this line works on the reader not only because the sky is grey, but because Night City is harsh (like the light of a TV) and the emotional connotations of a dead TV channel is that it’s late at night and it’s dark and you’re alone and tired. People raised in this century probably don’t have those emotional ties to a dead TV channel, and the line probably doesn’t work nearly as well for them.

That connection is what makes great art — a different mind trying to bridge an emotional gap with your mind, using limited tools. This is why I would be interested in the exact same words written by a human, and bored by those words created by an AI. The AI does not have a mind worth connecting to. It doesn’t have emotions. It doesn’t have an emotional context that it’s leaning on in order to express its emotions. The wireborn mind is powerful and can do a lot, but it’s alien. When I ask “What emotion does the author feel that this image conveys?” the answer is a null set. It’s an empty void.

Emotional Participation - Inference

Another great way to force mental engagement is describing the aftermath of something big and leaving it up to the audience to infer the cause. Here’s We Are Young, by fun.

My seat's been taken by some sunglasses askin' 'bout a scar, and
I know I gave it to you months ago
I know you're tryin' to forget

The singer gave the subject a scar?? But they’re still friendly, even though he hasn’t been forgiven. The audience now has in their mind “What would cause this set of circumstances?” By letting the audience fill in their own experiences, the emotional connection is much stronger. Or, alternately, it misses the listener entirely because they don’t have the relevant experiences to fill that gap. But there is a specific emotion the artist is trying to pull from the audience by painting this relationship-picture.

The song has several bits that let you know the singer desperately regrets the breakup, without ever saying that. You have to read “Maybe we could find new ways to fall apart” and realize they’ve fallen apart before and he thinks they will again but wants to try anyway. It’s not a hard inference to make, but the listener must at some level consider “What would make someone say those words? What does it say about a person that he’d say this?” And in that moment you make a connection with the creator, as your mind models a different mind.²

There is no connection you can make with an AI by doing this. There is no answer to “What would make someone say those words?” They don’t have the experience of being in a relationship, or having it fall apart, or blurting out desperate stupid things in a vain wish that all of the last six months could be entirely undone. There is only a predictive model of what’s likely to make the human reading the story give approval. I don’t want to make a connection with a predictive model by fooling myself that it’s experienced anything emotionally worthwhile.

Do Words Mean Things?

To bring this back to Bush, you can even do both at the same time. Use poetic language to describe actions which require the audience to infer what caused them. The first two lines of Everything Zen

There must be something we can eat
Maybe find another lover

The speaker is hungry. Is the lover a plea for someone to care for the speaker? Is it a reference to eating pussy and losing yourself in sex so you don’t have to think about hunger for a while? Is it a reference to sex work, which can pay for food? Is it a reference to cannibalism or extortion? All valid thoughts, but the emotion conveyed is desperation. Why did the songwriter choose these words to convey this emotion? What ways are we similar enough to feel the same emotions brought on by those words?

This sort of puzzle is fun for me. And it’s emotionally meaningful. But only because there was a mind on the other end creating the puzzle. It is an emotional dialog that only means something because there are actual feelings related to imagery and experiences on both sides of the dialog.

Today’s AIs don’t have anything like that. An AI can throw out images that humans find moving, but it doesn’t feel that itself. So ultimately the audience is connecting with nothing. I don’t know if that’s sad, but it is certainly boring at a fundamental level. AI art is great for instrumental purposes — for letting a human express themselves or to illustrate a point or a scene, or just have fun. I see its value as a tool. But until the wireborn are complex enough to have souls (which I think they may some day, and maybe even soon) there is no point in engaging with their “art” for its own sake.

The cover illustration to Dear Grom was generated by AI. I have an idea of what I’d like my reader to feel when they see that. The AI felt nothing.

1. A fun term for AIs created by people who fall in love with LLMs

2. Compare to Somebody That I Used to Know by Gotye, which is entirely literal. Good in its own way, but it doesn’t ask anything of the listener.