My friend’s been working very hard on a project for a while - writing, editing, researching, cutting, doing voiceovers. He played around with an AI writing tool called Lex to see what kind of output it’d give him on his topic. And it was… good. Somewhat similar to what he’d written about it:
It’s not really good, but it could almost pass for it if you don’t look too hard.
Let me tell you why that is very bad.
First I will say that generative AI can do pretty cool things. These just look great:
They can also be really funny - they have a gift for greentexts in particular:
And there are many other good and exciting potential uses for them. But when it comes to art specifically, I’ve disliked the AI generations in my gut since I first saw them, and I don’t think it’s just about fearing a world where human creativity is obsolete.
There are many common lines of critique that won’t be new to you and which I won’t cover here.
Instead I’ll focus on two things that I haven’t seen voiced elsewhere but which are I think importantly true:
the ongoing proxy arms race
the living energy in creation
I’ll start with the less weird one:
the ongoing proxy arms race
Part of why AI generators might not seem so bad is that their methods aren’t that different from what’s already happening.
Take Hollywood. From Sasha’s piece on Everything Everywhere All At Once:
I guarantee that [Hollywood execs] are going to do exactly the wrong thing like they always do. Sure, after this, the Daniels will probably get to make whatever they want to, which is fantastic. I am extremely happy for them. But also, writers like me will be called into Studio City offices, and told, forget the unique talent that got you here, we need something like Everything Everywhere.
But there will be nothing like that movie. No contrived attempt to generate something that riotously bizarre will work out. It’ll just be some hashed-up bullshit, with the kind of profundity/silliness feedback loop that EEAAO relies on, but without the deep structures of individual artistry that made it work in my new favorite movie. It’ll probably be totally awful.
Imitation is the minimal risk, minimal talent, maximal defensibility-to-management approach to success. And so it’s what rules in our world:
The problem is this deep structures of individual artistry thing. When you’re imitating success rather than originating it, you’re just replicating the outward forms and markers of whatever made the thing great. But it won’t ever be the real deal - it can’t be. The core inspiration isn’t there.
And so the key is just to fake it well enough that it passes. If you don’t have real quality, what you have to do is layer on all of the things people associate with it - its proxies - and hope your audience doesn’t catch the difference.
For example: a script can make up for weak dialogue by leaning into the rhythms and affectations of snappy banter, and done well enough people will often go along without noticing a difference. Or a story can play to the sympathetic narratives and currents of a time, things people are accepting more or less at face value, to create a false sense of meaning and depth.
People have always done this kind of bullshitting. Think of the way people try to seem smart or sophisticated - shoehorning a reference to the latest trendy book in conversation, or speaking with certain affected tones and postures. These things bear little on whether you’re actually smart, but people fall for this stuff all the time.
Associations and heuristics like these are around because they help us make decisions about an overwhelmingly complex world in real time. But they’re imperfect by their nature, and there’s advantage to be had if you can find ways to game their weak spots - at least until people catch on and update them. This is the proxy arms race: the complex, ever shifting battle between the ways bullshitters try to game us and our strategies for catching them.
There was a time when it was a fairer fight. But large, organized, well-resourced interests like Hollywood came in and systematized this game. They have spreadsheets for how to key into every target demographic to make their movie a hit; formulas and algorithms and sophisticated frameworks for how to make you cry, or laugh, or feel deep hatred for a villain and satisfaction when they’re thwarted. The term tearjerker is telling - there’s something about it that isn’t fully motivated and natural, but you find yourself crying anyway.
Things are kinda bad already. I’m wondering to what extent the pervasive sense of atomization and meaninglessness in our culture could actually be the result of this process happening along many dimensions - to be surrounded by so much deceptively compelling culture and art with only finite depth, none of it cohering into some deeper structure of meaning because none grew out of one.
It seems to me that AI will shift all of this into overdrive.
It’s because of the way it works, its creative process. Generative AI is pastiche, imitative, unseeing - mere recombination guided by statistics and powerful learning models based on trial and error. And because of this I think what it’ll end up doing is just to get better and better at gaming us, exploiting proxies for the qualities we are looking for because it can’t actually do the real thing.
Good art I think is ontological: about reacquainting you with the reality underneath your concepts, showing you a way of parsing the world that preserves more of its beauty. It’s a way of seeing and being that’s conveyed, the imprint of someone’s soul; and I suppose my deep belief is that it’s far too complex a thing to be arrived at by chance methods, even if guided by humans. Their input is just at too high a level.
And even supposing it could - would it not be 1000000x more likely that it’d just end up figuring out ways to game us?
While people are sensitive to the discrepancy between the real thing and its mere outward forms, I expect AI art will get old quickly:
But as these models advance, the approximation might get so good that even the most sensitive never consciously realize the difference; its hollowness instead registering somewhere beyond our awareness, some nagging feeling of emptiness that we can’t quite place.
None of this is to say that AI art can’t be interesting or useful. It just won’t be beautiful, or meaningful, or any of the other deep qualities people are drawn to in great art.
And that what we could get instead is a world full of creation that merely passes for beautiful - where things can be cool or interesting for a moment, but nothing has any enduring value. A world that looks good, and sounds good, and seems good, but just isn’t quite right in some hard to place way…
Now for the weird thing:
the living energy in creation
When I was younger I noticed in the car that when I’d look over at other drivers, they generally turned back - or at least seemed a bit disturbed and glanced around, even if the angles made it so that there’s no way they could have seen me.
I still do it sometimes, driving and in other contexts, and it’s uncanny - as if they’re feeling my attention through some sixth sense.
I don’t think it’s just me. People will say things like I could feel her eyes boring a hole in the back of my skull. Or the way people might think of each other at the same time and call - my ears were burning. There’s just some kind of strange energy involved in consciousness and attention, and people can often sense it coming from others even if they aren’t always aware of it.
I see confirmations of it everywhere:
There’s just something to this stuff.
So what might that mean for art?
I have a strong suspicion that this consciousness attention energy goes into the things people create. I don’t mean that it can merely influence the work and be seen in its subtleties - it’s more than that. I mean it’s imprinted there, directly, and registered by everyone who comes into contact with it whether they’re consciously sensitive to it or not. There’s life in the work.
It’s even true of something like a tweet, and it’s not just the subtle cues in the word choice or profile picture. It’s the person’s actual energy, the deep nuances of what they were thinking and feeling and being at the time, imprinted in the creation of the tweet and picked up as it enters your world.
And so you could have two works of art, perhaps sculptures, that are identical in every physical way. One is the original, the actual object that shared in the artists life and world; and the other an exact replica made by a machine.
The contention is that, on some level, the replica is not the same and would feel sterile in comparison.
I think it’s part of what’s behind the enduring mythos of a home-cooked meal. It’s not just that the ingredients or preparation or environment is better - it’s the energy put into it.
And maybe it’s also part of the appeal of live music. Not that recorded music doesn’t have this energy in it - it does - but it’s different. In a live show the artists are responsive to the energy of that moment in time, in that particular place and among that particular set of people; there’s a dynamic interaction between all of the elements, an energetic responsiveness and reciprocal amplification that can lead to really remarkable performances.
This energy is what I think people really yearn for most, to know and connect with others at these levels; and yet because it’s so subtle and requires such sensitivity to notice, we lose sight of it, instead pulled to think in terms of all the concrete, legible, repeatable, socially ratified things instead. And we forget what it’s like, or that it’s even possible, until something comes into our life and reminds us.
And so imagine if every little thing around you was a labor of love, full of the intention and care and spirit of another being:
Now imagine the opposite?
I am both impressed and deeply worried about the recent advancements in AI and people's reaction to that. The initial feeling was when I've started looking at AI artists and the bulk of work they produce. I was following specifically certain twitter accounts and was amazed from the frequent amount of high quality images produced, but I knew something feels very bad about it, knowing the people in particular wouldn't be able to produce one of these pictures if they had to manually draw (it would take me months and years in photoshop to even learn doing that) and also real artists won't be able to produce that much in such a sort period. It feels at first like something is lost here, it might seem pointless for artists to continue their work and people to appreciate art, if you just need some text prompts and several iteration even as a non-artist to produce that.
I can't predict what's going to happen. It's here to stay, but I am feeling similarly uncanny about it. I can't find the point when the process of creation is missing, but some people might do. I see something similar now with programming. While writing complete apps with AI is not there (it can write certain small algorithms, sometimes with mistakes, so you have to code review the results severely at times) I see articles popping up from professional programmers being delighted that the new age is here and one day "typing words to code" will be considered a thing of the past. And I wonder why this persists in this field? Programming is a form of art for those nerds who were self-taught and they write and share their own coding experiments in the communities. Maybe less so for the professional programmer? I don't get why people are delighted, maybe most don't like programming even if they are in this field (I meet sometimes people with CS degrees, confessing to me that they don't like programming).
My only hope is that sometimes people are seeking for skills to learn and things to build. Building stuff with your hands seem pointless when factories or a 3d printer can do it for you, but it's the creative process that gives meaning. If we reach the absolute point of pointlessness then we will see more people going back to the roots and building their own stuff, in a future when everyone will have forgotten how one does paint or writes code. (I am also thinking of that David Lynch video where he builds a wooden phone holder and delightfully presents the creative process)
But I don't know how this will hit the professional work (e.g. programming, artist or other jobs).