Email or username:

Password:

Forgot your password?
Top-level
Baldur Bjarnason

Generated art correlates strongly with low effort writing and video. But I keep seeing developer types in my feed reader use generated illustrations without seemingly realising how an increasing number of readers interpret it

And it’s only a matter of time until “generative art = spammy bullshit” will be the majority position because that’s how the economics of it are going to play out. Using extruded synthetic art will not do your writing or video any favours in the long run.

16 comments
CynAq 🤘

@baldur I reject articles with generated art not because of perceived low effort but on principle.

Side note; I love how everyone can instinctively spot generated text and art now. We came from “omg we’re all going to be blackmailed by fake photos” to “omg, enough with this fake bullshit” in a few months 😂

Baldur Bjarnason

@CynAq Yeah, most generated art has so so so many small tells

Priscilla Haring-Kuipers

@baldur isn't this the same for stock photos? Which would be the other option for people writing but not making visuals (like myself)

Baldur Bjarnason

@priscillaharing Stock photos don’t have as many tells. Between the colour scheme, upscaling artefacts, depth weirdness, text handling, off anatomy and the rest, generative art is kinda easy to spot, for the most part. Almost all of it has at least one of the tells out of the box

I also think stock photos require a bit more effort in general than generative art, for the most part

In any case, I’ve never seen people dismiss something I’ve linked to out of hand because it used stock photography

Baldur Bjarnason

@priscillaharing (I’d also argue that most posts don’t need generic visuals. If it has a specific purpose and is needed to complement the text, sure, but I’m sceptical about claims that generic images make any sort of difference to either traffic or reception. But that’s an entirely separate debate and not relevant here 🙂)

Priscilla Haring-Kuipers

@baldur visuals 'draw the eye' and most writers are required to deliver royalty free visuals with any article. I would expect a most fitting use case here where generated visuals replace other non-artistic, soulless visuals
Stock photos to me also indicates a lower class article as really good (well paid) writing is under a serious editorial team that would have some artists/photographers as well

Baldur Bjarnason

@priscillaharing Oh, I'm sure that writers are required to deliver visuals. I just think that the "draw the eye" statement for web-based articles is a superstition based on a small number of badly structured studies that were made years ago in a very very different environment, but that's not something the writers can do anything about so not really something that can be constructively debated

And again, generated art is becoming to stock photo what stock photo is to editorial photography.

Baldur Bjarnason

@priscillaharing Ah, sorry about the digression in the previous post. Still drinking my morning coffee.

The point I was trying to get across is that generated art, because of its statistical nature, tends to have a number of shared artifacts that are fairly easy to spot. Which makes them immediately counterproductive in any context where they appear as a preview thumbnail.

Petra van Cronenburg

@priscillaharing Normally, stock photos are photographed by humans and sometimes edited with software by humans (Paintshop, or making it less blurred etc.) It is only a special kind of selling photos or give them for free. Therefore, many seem to be soulless: You make enough money you have to sell masses.
But the boundaries are blurring because genAI photos are appearing more often there. In addition, some software or cameras already work with algorithms (for image sharpness, filters).
@baldur

Petra van Cronenburg

@priscillaharing 2/2 Nevertheless, such a photo can come from famous photographers. It's complex.
@baldur

harmonicarichard

@baldur That time has already come. I see plenty of anti-AI generated images on FB groups. (by plenty I mean two or three in recent days).

myrmepropagandist

@baldur I largely agree with this observation— however ‘AI style’ is a rapidly moving target— a target invested in *not* being distinctive or detectable. Right now AI-synthesized speech is robust enough that most of us could not reliably detect a generated voice speaking one or two words (although, given a full sentence, we might be successful) I’ve been fooled by writing especially for more technical articles. (see my ranting on online misinformation about ants)

Baldur Bjarnason

@futurebird Yeah. Though, unlike audio and text, generated images are still originally quite low res so almost all of it has that "oddly sharp" upscaling style even when it doesn't have any of the other tells.

Petra van Cronenburg

@baldur Unfortunately, as a journalist, I have to say that the falsification of photos also by AI is now so advanced that even trained professionals have to go to ever greater lengths to recognise it. It sometimes requires specialised technology. And media literacy in social media is not as high as one might wish.
People wouldn't cheat for a blog perhaps but the technology for misinformation works. @futurebird

Richard Barrell

@baldur on lobste.rs I usually now see most posts with a GAN hero image on them attract a couple downvotes marked "spam".

Go Up