The AI-generated art of DALL.E 2 - Robots with Flowers.
#1
A lot of you may have already stumbled upon this amazing new AI tool through Discord links or other posts from social media, but I still thought this was worth a share.

I read about a new AI tool called DALL.E 2 on James Gurney's blog, and I've got to say, I was completely blown away by the result:

http://gurneyjourney.blogspot.com/2022/0...owers.html

DALL.E 2 was used to generate a complete art book of Robots with flowers growing out of their head in numerous different styles. A lot of these artworks completely blow regular artists out of the water and could very well be used as concept art or inspirational paintings for animated movies, video games, and similar:

https://archive.org/details/1111101000-r...ew=theater

For now, it doesn't seem like this AI will be able to generate a storyboard for an animated movie or very realistic paintings, but it's certainly getting there. Thoughts on this technique? Will all concept artists be out of a job in time?

Reply
#2
Wow, i'm pretty impressed. They all look like something a human would have come up with at first glance. I definitely think this technology is going to start to be used at least for marketing, children's books maybe. If you need an image like these, why would you pay an artist to do it if you can just generate them? It seems useful for visual research as well, it puts different aesthetics together in interesting ways.

However, the AI just doesn't have a human mind. The art that is genuinely treasured, and not just an acceptable product, will always have a human touch, and human pathos. If someone came to me with some of these images and I thought a human had done them, i might critique them by calling the ideas vapid and well, robotic! There's a few that I would never have guessed were AI generated, and actually seem to be thought provoking like these:




However that is the human showing through with clever prompts. 

I think the fact that enough art like this has been made and posted that an AI can reproduce it might show that we will have to start thinking bigger. Maybe it will be a good thing if it makes our standards for art grow with the technology in the same way that the invention of the photograph didn't kill art, but sent it in a direction that looked for beauty away from mechanical accuracy. My natural reaction is to feel intimidated by this tech, it makes me understand the Luddites that tried to destroy new machinery. But new technology has always made us value great work done by hand even more. Now that chairs, tables, clothing, etc. are cheaply mass fabricated, how much more do we value a hand-carved piece, or a piece of really fine, tailored clothing? I think this makes me think a lot about what in art is really valuable and what may as well be generated by machine.

They are working on AI generated music as well, now which is pretty cool. It can look at thousands of scores and then kind of piece together something new that makes sense. But again it's like, it will probably get to the point where is sounds okay aesthetically, but it's no one's favorite song. People might even start to hate art and music that sounds computer generated.

Reply
#3
Some more images from James Gurney's blog that were fully generated by the AI:










Honestly, I can't wait for this software to become available! Playing around with it sounds like a LOT of fun!

I think that the biggest problem I can see right now is that the market will become extremely saturated with illustrations which will lower the value of illustrations in the future. Most book covers for example just use stock photos rather than illustrations nowadays (excluding children's/teens book covers) to save money. What's to say that an editor or writer won't just go with an image like this in the future and save on time and money? And this is also just a start to this software. I can already see it creating animations, full comics, and maybe even video games in the near future. We are in for quite a ride it seems ...

Reply
#4
The output is very impressive and I do think we'll be seeing more and more computer-generated art used in the place of hand-made art, not because it's better, but because it's good enough to justify not paying a skilled artisan to produce something more unique and tailored.

I think the main issue for AI-generated art for the foreseeable future is that it can only produce something derivative of styles that humans have created. DALL.E 2 looks like most of its input was 20th century American art (see the "painting by Caravaggio" in Gurney's post that looks more like a Michael Whelan piece combined with still life flowers). Obviously, this isn't a problem for commercial purposes, but it may mean that hand-made art will continue to have some value in the average person's, and hopefully average client's eyes. As far as concept art goes clients might still prefer a human that can follow instructions more precisely. Unless it's something like "design Sci-Fi soldier #15262462".

Joseph may be right that people will begin to hate computer-generated music and art. It's really a miracle when a piece of art is created that is widely enjoyed by people, and I doubt that anything made by a computer will be able to hit the right notes, based on what can be produced currently.
Reply
#5
(05-02-2022, 06:51 AM)Pubic Enemy Wrote: The output is very impressive and I do think we'll be seeing more and more computer-generated art used in the place of hand-made art, not because it's better, but because it's good enough to justify not paying a skilled artisan to produce something more unique and tailored.

I think the main issue for AI-generated art for the foreseeable future is that it can only produce something derivative of styles that humans have created. DALL.E 2 looks like most of its input was 20th century American art (see the "painting by Caravaggio" in Gurney's post that looks more like a Michael Whelan piece combined with still life flowers). Obviously, this isn't a problem for commercial purposes, but it may mean that hand-made art will continue to have some value in the average person's, and hopefully average client's eyes. As far as concept art goes clients might still prefer a human that can follow instructions more precisely. Unless it's something like "design Sci-Fi soldier #15262462".

Joseph may be right that people will begin to hate computer-generated music and art. It's really a miracle when a piece of art is created that is widely enjoyed by people, and I doubt that anything made by a computer will be able to hit the right notes, based on what can be produced currently.
Some human art is already pretty robotic specially with people with a more strong left brain type you might not even see the difference.What make art human like ?Being able to pick up on Cultural reference and using that to go viral?Imagination?Subject matter?The unique way you use your tool?

I see this more as a tool for illustrator but really do the Ai understand concept like composition?That would be hard to believe. How precise can it be... i certainly don't know.But i know that the more you feed info into an Ai the better it get overtime.

I don't think artist will stop being artist i actually think we will be seeing an over saturation and perhap the emergence of a trend toward human art vs Ai generate.

My Sketchbook
The journey of an artist truly begin when he can learn from everyone error.
Teamwork make your dream work.
Asking help is the key to growth.
Reply
#6
Those are good points, Darktiste.

I was thinking more about this topic the other day. Not very hard, but more than usual. Most art that is professionally produced is so stifled and derivative that it might as well have been produced by artificial intelligence. The same is true of modern popular music and even literature. We can thank risk-averse executives and the collectivist nature of modern organizations (in order words, THE MAN) for this. Professional artists working in movies and games often seem a bit depressed because of the robotic nature of what they're asked to produce, usually with a hell of a lot of 3D and photobashing. There aren't any tentpoles like the original Star Wars which was obviously dominated by an almost singular artistic vision (Ralph McQuarrie's, in this case).

So AI art that is 100% derivative and produced in a few seconds fits right into this part of the art world. But it's still iffy when it comes to rendering humans and animals, unless it produces something VERY similar to an existing man-made image. Renderings of natural landscapes and buildings are also often iffy, but this is less noticeable. It often produces images that just don't make physical sense (for a lot of examples of this, see art produced with Midjourney, which is already flooding social media despite only being in closed beta). An AI just can't "care" about these things in the way humans do, and it never will unless there is such a fundamental advancement in AI that it literally becomes conscious.

It will most likely improve in this area, maybe by a lot, but the big question is: will it get good enough for furries to visualize their fursonas with? Will a hand-made, individualized rendering of some guy's one-winged sparklewolf by someone like Reykat or Miles-DF remain something valued, maybe even seen as a marker of great taste and discernment? As Joseph wrote, markets for artisanal, hand-made products persist to some extent, even after a century of cheap mass production. AI art lacks everything that is valued in the most lasting and widely-loved works of art, except for usually giving the impression of virtuoso skill.

There is even apparently still an audience for human chess tournaments despite the fact that robots can play it better (albeit not because of anything resembling human-like intelligence). People can naturally detect the lack of emotional or philosophical significance in something that was generated by a computer, and often have an unhappy sense of being tricked if they happen to like something that was computer-generated but presented as man-made. Though the computer program itself is obviously an impressive human achievement. So I don't know what's gonna happen here, but it will be interesting to watch.


I am curious to read other people's thoughts if anyone feels like slapping the keyboard.
Reply
#7
(05-21-2022, 02:08 AM)Pubic Enemy Wrote: will it get good enough for furries to visualize their fursonas with?

The robots are takin' OUR JOOOOOOOBS!!

Reply
#8
Hahaha. I was only half-joking though. If you wanna start making money by drawing ASAP, you go for furry commissions.

Furries are the last domino.
Reply
#9
I guarantee I can draw a sexier wolf-man than an AI. No, I won't prove it.

Reply
#10
(05-21-2022, 11:15 AM)JosephCow Wrote: I guarantee I can draw a sexier wolf-man than an AI. No, I won't prove it.

Even if you would you would have to be on a waiting list to compare it at the moment.It still in experimental stage it seem...i suppose it because they have to train the ai not to generate certain keyword and behavior.

My Sketchbook
The journey of an artist truly begin when he can learn from everyone error.
Teamwork make your dream work.
Asking help is the key to growth.
Reply
#11
I have faith in you, Joseph.

(05-21-2022, 01:20 PM)darktiste Wrote: Even if you would you would have to be on a waiting list to compare it at the moment.It still in experimental stage it seem...i suppose it because they have to train the ai not to generate certain keyword and behavior.

This might be right. I read from one user of Midjourney that it tends to get dark and perverted in an ugly way even when inputting mild words like "couple". The creators banned many keywords to avoid the worst of this. Even the most sophisticated AI art generators are iffy when rendering humans right now, so all you would get is a grotesque melting human orgy anyway.
Reply


Forum Jump:


Users browsing this thread: 4 Guest(s)