[punkbit]

Sadly, some people believe that Generative AI output is the same; I find it very hard to watch and extremely cringe. I did see some good examples, but still not the same type of quality. For untrained eyes, all look the same, but I'm definitely not that type of person and very supportive of proper craftsmanship. You can see this same thing happening in software engineering and product design.

[socalgal2]

your comment makes me think of how photography was treated by painters. I think similar things will be true for generative AI. It will mostly be trash, just like most photograph is forgettable. But, a few people will find ways to use it to make amazing things, the way amazing photographers have elevated photography.

[jarjoura]

My only problem with generative art is

1: right now, it's basically 90% beautiful hyper-realistic woman in some scene.

That is SO incredibly boring. These models are able to generate all kinds of never before seen mixed together content. Yet, everyone just wants to generate photos of things you could have already searched for.

2: creative people should be the biggest and loudest voices celebrating it but the vision researchers building the tools have only been interested in solving the math side of it and haven't bothered to ask creatives what they even wanted. It's fine in private R&D, at the hypothesis stage, but then these corporate heads went and shipped as finished products and charge us serious money for it. So, creatives rightly feel uneasy about the models that legitimately "stole" their content without permission, while also minting new millionaires at these companies by selling it back to them.

Once companies license the content and artists can make royalties, and the tools to interface with the models become sophisticated so that you're not one-shot prompting into existence a random thing, more creative people will come back. (i.e. ComfyUI)

---

Plus, these models need actual fully staffed product teams behind them to solve and own the ethical issues around things like character identity locking that can be used to deceive. Or being too easy to say "in the style of" that could almost perfectly replicate another artist's entire voice. Then, there is the nudity issue. I don't want generative models to have morals, especially when it's art. Art is meant to be provocative and boundary pushing. I realize that's an entire deep-hole of shit in there waiting to get through, but that's why we need people spending their work day thinking about it more than just blocking anything that shows too much skin.

[perching_aix]

I'm sure some people believe that genAI outputs look the same as manmade ones. But for most, I think it's pretty clear that they just find them passable enough / not meaningfully different instead (or maybe even interesting in their own way). That is not the same thing: it's not that they don't notice, it's that they do notice but don't mind it. They're not being fooled, as there is no trick from their prespective. The differences were simply never a concern for them to begin with. There isn't even anything wrong with that per se [0].

I can also appreciate how a significant chunk of such images would make someone wince though, but I find it more than a bit confused to then bring up "proper craftsmanship" as the contrast. There's an incredible amount of extremely mediocre illustrations, anime, and other related work, that were made without any genAI tooling. High effort is reserved for the probable hits. Anime is especially teeming with extremely cookie cutter productions, and heavily overused tropes in saturated genres. I don't think the absolute chokehold that the high school romcom and isekai fantasy genres have on recent seasons needs much introduction. You can often just take one glance at a promotion visual and already identify both the genre and the quality right away. Cue the big 3DCG enemy on episode 1 (or 2).

It's also hard to reconcile this with the fundamental fact that the audience of these works is the mass market. Mass market targeting things are inherently bound by mass market sensibilities. Willingly not adopting a breakthrough technology - provided it is actually that - is about as far from proper craftsmanship in such context as it can be. So is being a one trick pony and sticking to an older fad. Making something good while locked away in an ivory tower is one thing. Making something good while simultaneously riding out the wave of actual real world circumstances is another [1]. I respect the latter a whole lot more than the former. Especially since the latter tends to devolve into self-abuse contests, to round off the self serving tendencies they inherently carried anyway. Code golf is a positive example for this. High art, maybe not so much.

I really don't see any reason to go on ideological adventures like this anyhow. You can just not like it, find it cringe, and leave it at that. It does not need some big philosophical undercurrent to suck. It can just look like garbage, be used lazily, and suck that way. And others can simply just disagree with those labels applying.

[0] A friend recently showed me their expensive sunglasses collection. They are indisputably expensive indeed. They are also the same cheap-feeling finnicky garbage as every other pair of sunglasses I've ever encountered. One can appreciate the expertise of someone who can recognize them as expensive, sure. But I think it says a whole lot more that the average person cannot. It's a sign of deep, long running cultural recursion, not unlike how when an LLM is fed LLM text, it degenerates.

[1] Could the average run of the mill show maybe look better and be more animated using genAI than they currently are? I definitely think they might just.