Instale o Steam
iniciar sessão
|
idioma
简体中文 (Chinês simplificado)
繁體中文 (Chinês tradicional)
日本語 (Japonês)
한국어 (Coreano)
ไทย (Tailandês)
Български (Búlgaro)
Čeština (Tcheco)
Dansk (Dinamarquês)
Deutsch (Alemão)
English (Inglês)
Español-España (Espanhol — Espanha)
Español-Latinoamérica (Espanhol — América Latina)
Ελληνικά (Grego)
Français (Francês)
Italiano (Italiano)
Bahasa Indonesia (Indonésio)
Magyar (Húngaro)
Nederlands (Holandês)
Norsk (Norueguês)
Polski (Polonês)
Português (Portugal)
Română (Romeno)
Русский (Russo)
Suomi (Finlandês)
Svenska (Sueco)
Türkçe (Turco)
Tiếng Việt (Vietnamita)
Українська (Ucraniano)
Relatar um problema com a tradução
Stable Diffusion 1.5 should work better but there wasn't a 1-click installer for it last time I checked. I'm using the 1-click installer for 1.4
Also it expects you to use an Nvidia graphics card. I'm using an AMD APU so mine is stuck in CPU only rendering mode. Each pic took 1-2 hours.
Someday I would like to build a rig optimized for running Stable Diffusion. But I lack the money for that.
As for the AI art missing something, that I feel is what real artists still have over AI art. There is always that artificial look when it comes to AI art, and I don't think that artificial look is going anywhere considering it's all made artificially to begin with and will remain that way.
That music video was crafted through the process of stitching together 10,000 frames of AI-generated imagery via Midjourney AI, linked together via individual prompts entered manually into the AI program. It uses a Machine Learning (ML) algorithm trained on a large amount of image data to produce unique images.
So yes, it's taking in thousands of reference images to create it's own. Such as doing a google image search upon the keyword(s) you wanted.
To create AI art, artists write algorithms not to follow a set of rules, but to “learn” a specific aesthetic by analyzing thousands of images. The style of other people's artwork, art techniques and brush strokes is taken into factor.
https://www.youtube.com/watch?v=m5FnWkSxNp8
I feel if the AI in question can become complex and nuanced enough that it contains elements of learning that in humans we would instead call "inspiration" from existing art styles, then not necessarily. If the AI remains primitive enough that you can look at its output and very clearly see that it by design copied several distinctive styles and simply produced them in combination, then yes I feel that represents an ethical (and potentially an IP) quandary.
If the AI gets to the point that it is simply doing something analogous to what human brains do though - learn art by initially copying an instructor or taking inspiration from other styles, usually early on by literally copying them - but then generates combinations of those other styles so complex as to simply appear at most inspired by them, then I don't see the difference in principle between what it does and what we do.
Other than intentionality, which at the moment AIs still lack, as far as we can discern. Which is where the theft argument becomes salient. If a user is telling it specifically, "Make something in the style of _____" and what it produces very clearly references their art, in an unvarnished way, then yes I think that has the potential to be a problem. Because that's intent.
But again, would we call it theft if the artist did this themselves and was homaging or taking inspiration from someone's style, without directly copying? I'm not sure. And if not, why does it become so if the AI does it? And at what point is it transformational enough in nature that it becomes fair use? That's another question.
So as usual, I think it depends.
I have no doubt it will improve. New versions of AI generators can already do photorealistic images. I think it's going to keep getting better, and keep getting more accessible.
There's already AI generators for pictures and text. There are projects working on AI music generators. I expect movie generators will eventually be a thing too.
Some day I expect they will take these generators and build them into other software packages.
Maybe someday someone will remake "rogue" but it will be a 3D game with all assets generated on the fly. The textures, the weapons, monsters, NPC's, quest text, music. It could all be generated on the fly by AI.
It's probably going to be really janky. But just like "rogue" was the first of its kind, this will be the first AI generated game and there will be a whole new genre of games following behind it.
I said this before but there is something special about human art, AI art always has this artificial look to it and that will always be the case because it is artificially made.
Also I'm certain that at some point the AI code that uses samples can output a list of all the sources it used and where it ended up in the final image. It would be stupid not to have that so the lawyers can cover their butts.
It's not stealing, it is AI learning from the images online however we still need to compile the image with Photoshop and/or algorithm. I think for those that don't understand it think it's just like Googling an image but it's more like a filter.
If people are not catching up with the times then they will fall behind. The same arguments happened with digital photography, printing press, computer based music mixing/composing and linen machines doing all the work. I'd love to see those that complain AI Art still use a film camera and traditional music instruments.