YHK - yet another hacker news clone

Everyone's sick of seeing the same default ChatGPT style on posters for everything from village fetes to coffee shops to school plays.

But I'm convinced that just vetoing AI is to throw away the baby with the bathwater. So I set out to prove it.

This feels like an AI trap almost. A lot of the badness comes from people who don't want/can't put effort into it, so we all get a deluge of the sameyness.

The author claims that the AI posters don’t have to be horrible, but fail to see that the dozen variations generated are not distinctive - they might have somewhat different typography or imaginary, but still have the same AI uniformity and smell as the crappy restaurant poster you see on the street. The core issue is not only technical, but also the lack of vision and imagination when creating art or even infographics.

  > I'm old enough to remember that people also complained about WordArt and Clipart and too many fonts etc.

  - those made by someone skilled/savvy (not necessarily a professional)
  - those made by someone without skills/knowledge

  > Also, it's just nostalgia

Designers republic and memphis were the only designs i couldn't immediately see were AI generated.

The only reason i couldn't tell memphis is because it's aping a known design language already, it's just a perfect copy so me not being able to tell it wasn't AI isn't a free pass in this case.

I think if there was a little less "grid" feeling to the layout details then i wouldn't be able to tell on a few of the other designs either.

Edit: the designers republic version is quite a lot different than the others, it's not just that it's the only design to avoid pairing every feature with an icon, it's the whole layout variety - there's different font sizes, the positioning of elements is very well done to my eye. I'm gonna say this has to be another case like memphis, there's some pre-existing design language that i don't know of and it's just aping that. It's much more human than the others which i can believe are straight ai.

But the thing is, all have that icon+text style in the "tombola, cakes and drinks..." part. If I wanted to be distinctive, this is what I would have tried to get rid of at first, not the "visual style" (it's sort of like Windows XP visual styles, no matter how different they are the overall layout is still recognisably Windows).

It’s not complicated - the default style signals low effort. Why should I get excited about the village fete if the organizers put in the barest minimum of effort?

But there’s another reason it drives people crazy - it’s low effort trying to present as high effort. If the sign was simple Comic Sans then it would at least be charmingly honest.

That icon+text row is the tell for me too, though I'd blame the prompt more than the model, nobody ever asks for an awkward layout.

That default style is very recognizable. I am seeing it everywhere.

I guess it is like the early days of wordprocessing, where you saw Comic Sans everywhere.

But if a designer improves an AI design and I feel it's fine, does that make me strange? If you ran a blind test on this, I think it would come out with high preference...

I think this is a misdiagnosis of the problem. The AI posters are just fine, perhaps even pretty good.

The problem is that using AI to create the posters show a lack of willingness to put in effort, and it makes you question whether the producer cares about the product and whether they care about the large scale societal problems related to the usage of post-2022 transformer based neural networks.

What's interesting is that you can't do the same with long-form AI prose. You can ask any given LLM to write in different styles, but it will always fall back on the same rhetorical and structural techniques, so the "new style" is strictly surface-level. In fact, it's getting worse over time -- especially in Claude from Opus 4.6 --> Opus 5.

I suspect that the same thing is happening here -- as other comments note: lack of distinctiveness, uniformity -- but that the low information content of posters prevents it from becoming immediately obvious.

AI posters may or may not have to be horrible, but they absolutely must include [ icon | icon | icon | icon ], apparently

I see, across my city and others I visit, every bar/restaurant/etc now has low effort one shot posters, menus, etc, some with old chatgpt nonsense text / errors. It's everywhere. All that brown-ish hue food 'pictures' etc. I mean 'authentic Pizza napolitana!' and then an AI generated picture of said pizza... One of them (an Italy born Italian restaurant owner here) and he just said; people come in with the very expensive pictures / menus I had before and the same amount of people come in with these free ones. And he actually likes these better as photographing (real, not mcd/burgerking plastic props with poison paint to make it look better) food is hard, even when paid for.

AI posters are not horrible, they just expose the person who "commissioned" them as trash. In a way, that is a good thing, since you can avoid them more easily.

Their very first attempt is arguably the best one there. I was surprised they didn't like it much and kept going. Then the "Stamp / Letterpress Style (Modernised)" was the only other good one.

So of the two designs made, only two were good, the rest was quite bad and lost all meaning.

Moral of the story - have taste.

Just like in engineering.

these are definitely better.

I really wonder what's going on in the minds of those who make the shitty AI posters? like, are they that unaware?

for me when I see them I just get instantly repulsed

Most AI-generated art is "good enough" for most one-shot designs by now (though still not [always] as good as a good meat-based artist of course)

The problems right now are:

- Maintaining a consistent and coherent style/aesthetic when you have to generate lots of assets for something. For example sprites in a video game.

- Iterating over a design for more than a few passes. For some reason most models keep degrading the entire image if you keep asking for edits to a specific part of the image. Or, some problems are or become impossible for them to fix no matter how you prompt them.

I remember seeing that default style long before AI. Probably where it came from. At least the default is not Corporate Memphis.

It's not that the first one is so awful in isolation, it's that they are everywhere you look when you walk down the street. The barbers, the school fête, the cafe, the church. Everywhere.

As usual, AI is a tool and will be good in the hands of the wielder. I've been saying this since Stable Diffusion released nearly half a decade ago.

I like the other posters more, but the first poster probably still conveys the idea of the event better: people having fun, drinking beer, with a live band. You don't get that from most of the other posters unless you first mentally transform those icons into real things, which takes effort.

People dont want to write multiple prompts or write a lengthy one. They want to say create poster x for event y. That's it, that's what they get and that's what they take.

Kids are rediscovering typography and pictograms. With AI. Well done!

Buried at the end of the article the author linked a library of different art styles applied to the same poster content. Found some new cool art styles to try out in the future. https://john.hartnup.uk/poster-prompts/

> They’re OK

No, they aren't.

AI is a tool, just like photoshop was. Just because someone bought photoshop didn't mean that they can magically create amazing digital artwork, they had to learn photoshop and also be creative. Otherwise it would have been a horrible poster with comic sans. Same goes for AI, one has to learn to use the tool and be creative.

> They’re OK

They are not OK.

I love how the AI silently injected "A day of fun for all the family" on the hand-drawn by a child design.

PS: Which then survives in the next iterations as "A day of music, making and family fun"

Edit: Ah just saw the call out on the previous iteration.

But still it's a nice display about the semantics encoded in the models.

Can you tell your AI to not apply piss filter to everything?

One of the biggest problems is actually making sure the information needed is on the poster.

Times, dates, location, pricing, parking etc…

As he points out “it added a bunch of text…” now it’s one all the future designs. That’s great - is the text correct?

To be honest this was a problem before AI. People are not very good at communicating.

I think we're in the 'early digital camera' phase of AI art gen.

The real photographers knew they were shit photos. They were grainy, bad lenses, and ugly. A lot of people still liked them because they were much more convenient for a lot of uses than the film ones.

Well, I guess there's still a lot of film people. But film had a ceiling. Digital cameras kept getting better.

The adaptive film photographers eventually adopted the tool. A lot of their skills transferred well to the new medium. Most waited until the cameras got good enough, though.

I'm not sure when that'll be with AI art gen. Seeing scenes broken up into layers in the latest gpt image generation, though, made me think we might be near.

At least the AI generated posters are better than human-generated t-shirt designs, where you find generic texts such as: "Est. 1984" (or any random year), "I [heart] NY", "Property of [City/School] Athletic Dept.", "Original Trademark", "Authentic Denim Brand", etc., etc.

All trash, in my opinion. Different flavors of trash.

At a certain point, pottery that looked like a child made it became the high end, because well-executed pottery was coming out of machines.

We've officially reached that point with advertisements. Save your tokens. Hire your 9-year-old niece to draw the poster by hand and you'll likely get a far better reaction from most people. Nobody goes to a fair because the poster had a "modernist / Bauhaus-influenced graphic style".

if NOT TRULY personal preference, then all those variations ARE horrible, but if they serve some taste--personal or marketing--then they are likely not to everyone's taste.

to me, they all look like lazy-prompting, which is the only way I ever do genAI imaging, because I fail to see why that "CPU" should be wasted on something I barely even care for.

That said ... let kids do that shit and pay them 50 or 100 bucks ...

The issue is that in my walled garden, I have not seen any of the good AI stuff. My wardens still use some home trained GPT 3.5 for most stuff, including their songs on potify and bootube. And i can't get access to current top level public AIs, either. Had the actual ChatGPT and Claude models for some while but I knew why ... otherwise I might put effort into semantics, syntax, grammar and my image generation prompts, which currently look like some MITM is speedrunning collages using haiku cowork and crayons or something ...

If only there was a way to precisely control our output so it matches our taste and imagination. I wonder what that looks like, maybe it will be part of the next generation of art.

I find it kinda funny that even "smartest", most capable models often have a very hard time going beyond surface-level, top-of-mind associations, when faced with creative tasks.

Of course, a "Japanese Minimal Poster" has a sakura and a stylized flag of Japan, duh

A human designer would probably think that "Japan -> sakura" and "Japan -> Japanese flag" are way too obvious, too banal, too stereotypical, and, most importantly too boring. And then they would probably sit for a while and try to think of something fresher and less clichéd.

But AI has absolutely no problem with going with the cheesiest, most overused trope.

Is there a way to formalize the space of all possible aesthetics a model knows, in a way that you could randomly sample from it?

The analogy in my mind when I read this post was: you are asking a machine to paint a color, but the machine always paints the exact same shade, deterministically; and you'd like to explore different colors, but somehow you don't know anything about colors, don't know any of their names. So you don't have the language to ask for what you can't name.

Random sampling of colors (without names!) is a solved problem, because we've parameterized the entire universe of colors that can be perceived. It's just a very small three-dimension manifold, and there's even a simple metric on it such that unit-distances in that metric are what humans say are color differences of equal step size.

Does ML have an abstraction for this? "Here's a list of every human art in history: create a coordinate system that uniformly samples every art that can ever exist?"

Adding a safety pin to the punk design was too much of a cliché.

Well, sometimes AI generates pictures that are not bad. Personally though I am tired of AI slop in general, and I don't want to support it. AI literally kills humans. In some scenarios AI is still useful (e. g. if you write a game, say, in javascript and you don't have any designers helping you out, you could generate images that don't suck, quickly, via AI) but other than that the less I get of AI the better. Google now has that problem with AI - tons of AI slop and fake videos. I realised the AI slop just wastes my time but Google does not offer a simple "do not show me any AI slop" videos. After all they want people to waste their time on youtube, by design. So I now waste my time twice by using Youtube, if I watch an AI slop video.

This isn’t a web UI, it doesn’t need icons.

But more importantly, if you aren’t the kind of person who could make a halfway decent text only poster before, you are doing to make the posters at the top of the article. Which are objectively harder to read than something typed up in word.

You aren’t going to know to tell the AI to “ Go for a clean, unfussy, bright layout with a bold striking spring-themed graphic. Avoid pastel/airbrush/oil style art or images of people.”