[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/ic/ - Artwork/Critique

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: file.png (257 KB, 1219x691)
257 KB
257 KB PNG
Honestly what i believe is a fundamental failure of AI that i dont think will ever get fixed is that since it's not sentient it only ever short of "predicts" what makes sense so you can never trust it to make anything really specific or say, fill out a thing and keep every part of it coherent with what is next to it, like if you want to fill out a children's book or even make an animation you cant really prevent the ai from saying fuck you and picking out a random part to exaggerate uncontrollably. Then you have to adjust either through prompting or manually at wich point you're better off making it yourself
>>
File: file.png (916 KB, 740x782)
916 KB
916 KB PNG
>>7242792
like here it doesnt really have any fucking clue what a "7th century iberian umayyad armor" is so it just vaguely follows "fat man" and then plasters the armor however it feels like, you cant really trust it to make it ride a horse in the position you want, nor make it be the same person or horse consistently
>>
maybe it'll get fixed but by how i understand language models it would have to also learn to analyse 3d models and then you have to build a 3d rig and then there's more work work work instead of being a tool wjere you "cheat" and you press a button and then it magically makes the whole comic
>>
AI fundamentally can't do perspective. It's like toxicity in LLMs no amount of training has fixed the issue, and belief it can be fixed is one based on faith not reality at this point.

The technology will have to radically change to fix these types of issues. Until then no amount of rendering will make up for this amateur quality (just like with beg artwork vs int and pro).
>>
Skill issue.
>>
>>7242792
AI can't even do the exact same picture twice with small corrections like changing hair and eye color, or adding make up to a character.
And that's the real reason why it's useless for real studio work. 90% of studio work is doing a million iterations of the same picture for "idea guys" who don't know what they want.
>>
>>7242792
yes AI is very bad at details and it'll probably remain so forever
it fails spectacularly at creating small things like accessories and texture because all it can do is create the impression of detail, not actual detail
>>
>>7242792
I think the technology will progress no matter what we do, but the problem lies in the human operation. I made AI hentai for a bit to try to earn some money. I only did it for a month and I already had started making some money back, but it was so psychologically damaging I had to stop at the end of that very month. When I went to sleep I couldn't think of anything but the prompts, what people would like, and all sorts of heinous thing like a poison.
I made whatever I had free and fleed.

To use AI for art is soul-crushing. It is no art, only a tool insofar as the corporate inhuman needs of its operator; for advertisements, for porn, anything that has a distinct use case.

I cannot define art, but certainly no AI fits in the subconscious representation.
>>
>>7247043
Because AI completely removes your agency, it's not only because it can not generate exactly what you want, and you need to fight every step of the way, but also because what it creates is its creation not yours, even if you manage to make it do what you want the process is so detached it feels inorganic, so you can't relate to what you created. Imo even if in some future an actual AI is created that can hijack your mind and deliver your idea at the mere thought of it, it still won't solve this issue, because art is as much about process as it is about result.

Artist flocked en masses to digital, because while not the same as trad it still shares the same process in principle, AI does not on a very fundamental level.
>>
>>7247048
Right, and even still, traditional art has its own appeal. Maybe it's something about the connection with the lines, maybe it's about the object really 'existing.'

Like a rock climber scaling a cliff without any safety gear is much more of a celebrated achievement than with safety gear.

In that intrinsic sense, the integrity of art will always remain intact, at least in judgement.
>>
Alpaca just updated their render anything trailer.
It removes the blocking to render part allowing you to correct and detail.
There will be a decrease 500 to 50 or 5 artists on projects.(DreamWorks)
>>
>>7247055
>render anything trailer.
so this is what commercial aesthetics will be looking like from now on and anons think capeshit art stagnation is bad, kek
>>
File: file.jpg (40 KB, 640x480)
40 KB
40 KB JPG
>>7242792
Oh my fucking god you brain dead imbecile stop posting these shit takes if you have zero clue how this shit actually works. It's obvious you're just barfing up whatever techbro marketing material says.

During training these systems literally just go through a bunch of tagged images and derive nebulous coefficients for keywords. When it's generating something it starts with random noise and uses a function that amplifies patterns based on keywords you requested. That's literally it. All the boxes shown in brochures are just there to improve performance, the principle is the same.
When you enter "fat man sitting" the model will definitely have a lot of sitting examples, so it will amplify a sitting figure. The face of a sitting person will be averaged out, and won't be one of a man, but since man is requested as another keyword it will be amplified separately. Likewise with fat. Or any other keyword.
Pic related, just averaged a few photos of a face. Didn't bother with scaling but even then you can make out a nose. The GPUs in some datacenter make billions of these per second and iteratively shit out one that fulfills the task with least deviation.

Stop clapping for this crap you vacuumhead. This is just images off the Internet photobashed and offered back to you. It has nothing to do with "intelligence".
>>
Is this a Cris thread?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.