Chen did it again @broeckchen - Tumblr Blog

A Hypothesis about Sending, Receiving, Touching and AI - Brückchen

Okay, okay, so here's a thought that coalesced in my brain today:

Let's say we're speaking about an image. To communicate an image - that is, to put a visual into the mind of another person - there is a minimum amount of cognitive and physical labor that has to be done. That distance is the Gap and the labor build the Bridge.

S |----------| R

The Basics

The Sender of the image can simply write or say the words "a cat". the Recipient imagines a cat. This is the Sender only putting in a small amount of the labor to put the image in the Recipient's head. S --|--------| R

To get an image of a cat, the Recipient has to do the rest. This can mean pulling the Sender closer through questions about what the cat looks like. Or simply imagining the details of the cat according to their own thinking. Either the Sender is dragged into more labor, or the image is likely inaccurate to the Sender's intention in some way as a result. S --||-------- R

This is what we often do in daily conversation, right? But that's okay because we usually don't need suuuuuper much detail to effectively manage to stay on roughly the same page. And often we share enough experience that when I tell my roommate about the kitchen, we think of the same thing anyways.

Then the other thing the Sender can do is put in more of the labor. For example, give a detailed description of the cat. S ------|----| R

The Recipient will still need to put in some of their own cognitive labor, and notably, maybe even a smidge more physical labor to read. It's possible that the overall labor put in increases as a result, but as you can see, each party still needs to Bridge a smaller Gap than the Recipient had to deal with alone before: S ------||----- R

And the resulting mental image is more accurate to what the Sender wanted.

The Sender could also utilize a completely different skill. Drawing, sculpting, collaging, photographing, etc... to basically take on almost the entire workload of this. S ---------|-| R

With an actual visual - like a picture or a sculpture - there is barely any inaccuracy left. And looking directly at the image as the Sender provided it is minimal labor on the Recipient's part. S ---------||- R

Here's the thing that's important to understand: Even if a drawing only takes someone 10 minutes, it's still more labor than a written description, and vitally, a larger amount of labor the Sender voluntarily takes on in this process. Writing is a skill. Creating a visual is a distinct, second skill. Chances are that the Sender had to learn both skills to a point, even if they never honed writing as much as the visual skill. So... extra labor put in. And vitally, radically fewer amounts of labor requested of the Recipient. I hope you're still all with me.

The Touch

So here is where I think GenAI rubs many of us instinctively the wrong way:

This communication process is inherently somewhat intimate, isn't it? It's very... human. An author going to great lengths to use the right wording is doing their best to invoke exactly the right mental images in us. A therapist rewording your burdens to hold up a mirror is applying a lot of personal experience and studies to make sure both of you really meet each other on that little ---------- line I typed there. And both Sender and Recipient shape their parts of the covered distance through their own lenses and abilities.

That's literally connection. Both sides reach out with a part of themselves so those parts can Touch somewhere.

GenAI is literally developed to slip into exactly that gap where we were supposed to Touch.

An image: AI S --||-------||- R

The Sender can give the AI the words "a cat". It will cover the distance between that and the rendered visual for the Sender and the Recipient. By doing so, it removes a lot of necessary labor from both sides.

Why that feels wrong

God I hate doing subheaders like this because AI does it now but they're really useful for readers who find walls of text overwhelming

Our first problem is that GenAI is trained on the vast body of human creative and scientific work. When it bridges that gap, for one, it puts in unsanctioned labor by those humans. Obviously, that feels instinctively wrong to us. We are used to associate labor with result, and we all balk at the feeling of labor having been done without result/reward AND results poofing into existence without labor being done. At the same time.

But also, and I feel like this is what we should actually focus on: We are eliminating this point where we... Touch.

I think it's important to focus on that Touch because that's more useful in literally... any nuance here.

When you read a book, usually you want to Touch the author's inner world. And you want for them to Touch yours. Chances are an editor is also in there somewhere, so you're having that Touch with more people than just the author. But still!!!

The moment GenAI writes the book, that Touch gets removed. Both the author and you touch the AI instead. In the reader's case, that tends to be especially frustrating because remember, putting in the effort to attentively read and imagine is labor requested from the Recipient! So you find yourself in something like this: AI S --||----||----- R

Where you as the reader have literally put in more labor than the author (or at least feel like you have, because you can never know how much or little AI was used so you can't pinpoint the labor amounts). Just to be left hanging without the Touch. Even if the author only used AI in minimal amounts, you don't know securely where. Which words are where you Touch, and which ones are where you might feel like you do, but in reality a robot is the one on the other side?

The AI averaged out a ton of the most popular written things and put it in the middle, and those things are neither the author nor you.

Nuance

Okay, so here is where this model gets useful: It helps us understand the disconnect with people who oppose the rejection of AI, and it might help to at least bridge the gap in understanding with those of them we treasure. Family members, friends, etc.

For one, it gives us nuance. It lets us distinguish between functional communication and expressive or connective communication.

A quick work memo is functional communication. You are trying to share a quick concept, usually. You're not trying to Touch. Hence why people who use AI to write work memos and emails often don't understand what artists are going on about when we speak of emptiness or lack of intention. The average of the most popular work memos humanity produced is a suitable substitute for the real thing, sometimes even an improvement to it, because it bridges the gap well enough and the absence of the Touch doesn't really matter here.

A love letter NEEDS the Touch. It's literally intended to create that Touch! Same with, say, wedding vows. These are things where the entire point is to Touch, to connect, to reach out for each other. Skipping any of that labor inherently breaks the entire communication part down.

This may be a controversial take, but I think a lot of what we colloquially call art sits in between these two ends of the scale. Especially corporate art. And I think this is why we feel the negative impact on the market so hard as artists.

The Luddites were right but also wrong

Corporations usually don't give a shit about the Touch and never have. They want to put minimal labor into the Thing and still have it reach the Recipient as a Product. To them, this gap was ALWAYS bridged by a third party. Doesn't matter whether that third party is AI or a human artist to them.

So you can't appeal to corporations or anyone capitalism-minded with this. That's why they often call opponents of AI Luddites - the word is genuinely fitting. And I mean that both positively and negatively. The historical Luddites sabotaged machinery because their work being automated led to less employment for them, threatening their livelihood.

The negative part of that comparison is that neither the Luddites nor many AI-opponents actually aim at anything but the machinery. The technology exists now. The genie has left the bottle. We can not turn time back. But we ought to seriously raise our gaze and look at the people who refuse to make this benefit all of us.

Like, am I the only one whose dreamjob never was to churn out cheap knockoff cereal mascots for brandless boxes? Probably not. Someone probably did that to pay the bills, while sitting on a graphic novel idea they never had time for. Their Touch with their Recipients was always distorted and barely there, because it was always a corporate idea they had to connect to the target market. It got filtered through them, but they were a middleman from the start. This is literally the Alienation of Labor Marx already wrote about.

That artist actually deserves to simply get a living wage even if their labor is not scarce. And they deserve to thus free up the time and energy to create things with which they can actually experience that Touch.

We might want to do better than the Luddites and throw ourselves into activism that demands the support structure we actually all need as labor gets more automated across professions. Or eat the rich.

Our Labor is invisible to Techbros, too

With this model of Sender, Recipient and Touch, you might also be able to understand the people who cling to this technology. To many of them, it was literally never a concept they understood that this Gap can be bridged through labor because skills can be built. So they just perceived themselves as stuck without the Touch because they thought they couldn't walk that distance by themselves.

Now they have a bridging tool. That feels amazing! They don't understand the immense cognitive labor put into covering that distance even before the physical part begins. They don't see and therefore literally don't know about the way many artists plan compositions and poses and many authors track the positions of characters in a room to get things right. Because the thing about this labor is that if it's done really well, to most people it becomes invisible. It only gets noticed where we fuck up.

Obviously they cling and defend. You're trying to tell them that they're bad and lazy for using this cool magic wand that allows them to get further on that road, while insisting it's bad because it replaces stuff they've never heard of before.

And the worst part is... I don't fully disagree with them.

Okay, hear me out, please please please!!!

The question here is, again, whether their images for example are supposed to be functional or communicative, isn't it? And how long that road is.

If someone who wants to commission me lacks both the writing skill and drawing skill to provide me with fully helpful references, but whips up something "close enough if you change x, y and z" in GenAI, that's very functional. It's literally just intended to help bridge the gap so I can create actual art that gets the full vision right. And commissions are a great example here because that's already how they work. They are a way to loop in a middleman to build the bridge from Sender to Recipient to a point. To create a visual readers or friends or other artists can use to have the correct character image in their mind.

So... is the person ordering the commission obligated to put in the extra labor, especially for the first reference piece? The idea behind the reference sheets especially is to communicate something, but I'd say it varies wildly how much of that communication is supposed to be functional vs connective.

Other examples I can think of: An artist wanting alt text for their images for the visually impaired. Self-written alt text is obviously best. But do we expect every visual artist who wants to provide the accessibility to hone their skill in that? Is that feasible? Do they have the time, energy, spoons for that?

My mother sent me a little AI generated video of herself riding a huge bunny for Easter. She's an office worker. That's a skillset she's honed all her life. Should she have to learn an entirely new one to create this visual?

She doesn't technically NEED it. None of the examples above NEED their visuals or texts. The commissioner can just go on without the art piece, the artist can simply live without visually impaired audience members, my Mom can just say Happy Easter instead.

I think it's super worth to argue that when it comes to the environmental cost of AI. The environmental cost is not really worth it for something we don't need.

Just artistically, I get why the "but you didn't make it" argument may not make sense to people in those situations.

But I think the nice thing here is, now you have a mental model that allows you to explain why the "you didn't make it" matters. Why there is so much value in what humans put in in the first place. In personal expression and communication, we seek and need that Touch.

Quick Aside: we do it too

A reason why arguing from intention also sometimes annoys me is because human artists are not always super intentional about our art either. As someone who has been around the DeviantArt circuit, sometimes I just made the shading this particular way to earn points in an RP group. Sometimes I chose an outside scene because it was easier to draw. Sometimes I did this super swanky frog view fisheye lense angly because a ceiling has less detail than the rest of the room and I just figured out a ruler tool for fish eye perspective.

I think arguing from intention with art in general can be... intimidating to young artists, or artists who have low self-confidence. I bet some people reading this have seen the argument about intention in art pieces before and felt a bit uncomfortable because they don't put that kind of thought into their average stuff. But I feel this is also why the Brückchen model is useful. Because the things you put into a piece intentionally, they carry your labor because you actively thought about them. The things you simply tossed in because [reasons] ALSO carry your labor, just different kinds. Labor like getting particularly good at drawing trees over anything else. Labor like knowing your strengths. And also, they carry what you found fun, or where your mind went first, and so much of the media that shaped you.

So it's still you reaching over that Gap, building that Bridge, and the Recipient still gets to have that Touch with YOU. Intention is just one way this can happen.

Humans like doing Stuff

Third thing this model is useful for might be even more unpleasant to hear about, but it's important to know of when we want to talk to people who are ProAI: The gap the AI bridges isn't always the same size.

Sometimes, a Sender gives the prompt "a sexy catgirl sitting on a chair." and our model looks like so: AI S --||-------||- R

But the interesting this is that this distance between Sender and Recipient isn't invisible to all Senders. Many notice it and want to make it smaller, to get as close as possible to the Touch. So they start getting really into it. They learn how prompts work in detail. How GenAI "thinks". What Words of Power* are.

They begin doing labor. And write prompts with a million specifications and codified bits and bobs and then iterate towards the closest thing to their idea. AI S ----||-----||- R

And the wildest part is that some AI-users, as Senders, start wrapping around entirely. They learn to sculpt in Blender - maybe not with textures and lights, but enough to create a rough 3D environment. They learn to rig and pose 3D models in that environment. They then pull the render into art software and learn how to block out the colors correctly, might even vaguely shade things, add details they really want in there. Install custom software that lets them tweak in insane detail where every single bit comes from, that lets the AI ingest several different reference images and compose them anew. They learn color theory and composition to get those right in that draft. They learn prompt engineering to then feed the composition draft into the AI with a detailed description.

Where does that put them?

Here? AI S ------||---||- R

Here? AI S --------||-||- R

All I know is that it's not here: AI S --||-------||- R

Do we consider that a sort of Touch? Almost-Touch? Like, what is that? I've only met a handful of people this passionate about it, but they exist and they are... somewhere. They're doing labor. They're learning skills. I suppose if we go by the 80-20 rule, they only learn the 20% that go the first 80% of the way to a finished thing.

But also, I am super fond of these people. Because I like people who get handed a technology intended to do the Thing for them and promptly find new forms of labor to put into the Thing instead. I think that says something hopeful about humans and our eagerness to keep going even when we are handed a magic wand. I think this is another point we could slot into and encourage: the learning around the new technology. The way even the magic wand gets so, so much better if you learn about the Old Craft. Maybe we can get people hooked.

Disclosure

As for me, full disclosure: I utilize AI as a tool. The topic is insanely interesting to me and I learn the most by actually grabbing things and taking them apart so yeah. At the same time, I am an artist so... of course I don't wholeheartedly endorse it. My feelings have always been pretty complicated here. I will never knowingly post any AI images on this blog masquerading as real art. I don't consider AI images art, even the ones with a lot of labor behind them. To me, actual art is very much about that Touch.

I participate in the AIRP community while also agreeing that the environmental factors are deeply concerning, so I look into ways to reduce my contribution to that cost. I do think human experts are literally always better and faster at everything AI can do. Including RP. I understand that this is still relevant bias to the topic, so I want to be transparent.

What if Humans

This post is not intended to sway people towards AI. Quite the opposite. I think it gives us a better idea of how AI damages our communication if we let it, and why it isolates us even when what we do with it isn't necessarily inherently social. And furthermore, I think this model of communication also lets us understand other humans better.

If you have a conflict with a human friend, in a good and healthy friendship both friends will put in the effort to make sure they truly Touch. If your significant other is trying to tell you something hurt them, the labor expands to include things like emotional regulation, self-reflection, in some cases even actual emotional hurt on the Recipient's end.

Rejection Sensitivity for example adds to the labor a Recipient of criticism needs to put in for the connection.

You can observe how much of that distance you are usually covering. How much of your labor goes into wording things compassionately? How much effort do you put into choosing this gift? And how much are other people willing to put in on their own end to reach you? Do you even overshoot sometimes in your effort to touch and kind of end up at something the Sender wasn't even Sending?

Idk. Call it Brückchen Model in my honor or dishonor or something. Brueckchen. Heh. I like Brückchen better because it has a lil smiley face in it: ü Thank you for coming to my TED Talk. *Words of Power are a concept in AIRP circles where certain words have such distinctively similar training data behind them that using them in a prompt will always put in entire clusters of behavioral traits. For example, "dominant" in a bot's definition will almost always make them aggressive and possessive by default, because that's how dominance is usually portrayed in media. These exist across prompting though, and give you a lot of interesting insight into people and the nuances of words we use.

Trending Blogs

Recently Viewed Blogs

Chen did it again