Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.Daz 3D is part of
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Let's move on with the OP's original topic and not turn this into a debate and comparison game between AI and NFT, please.
...creating "art with text" is nothing new, I used to do it back in the days of dot matrix printers. Wasn't as refined as say drawing or painting. Same for creating technical drawings on a pen plotter which I've done (also did 3d wire frames) . Even in the early days of 3D graphics it required manipulating text though coding. Still there was Aa good deal of personal effort expended in all of these. I
t wasn't simply pressing a key and "voila!" you have art.
As to "toasters" creating "art", the Sci-Fi series Babylon5 employed the Amiga Video Toaster for generating all the space scenes including ships, planets, and the station itself (loved those "spidery" Shadow ships).
Amiga... i went for the crisp black and white image, instead.
For a positive application, with no need for copyrightability, but the potential to spice up a social network, i would throw in "instant memes". So you might have some kind of relating to each other with posts, images, videos, an of course there is more of an interface needed, but the ability of such a system to create stuff in real-time as well as to relate to given images, could be crafted into a feature of a social network in theory. It may be burning some money, but it's at least something potentially interesting. Memes often live of images from movies or reality, however think of combining such somehow with a "battle mode", where people compete for, well nothing but the fun of it. [Then reconsider hate imagery with generators - so no sex, no Auschwitz, no ...i can see new EU-Law demanding output filters for all images ever made enywhere.]
So, combined with interfaces, and perhaps some advancements, there really is "no limits" to at least interesting applications. Of course training is expensive...
?*HMHbb__ ```*#HM#b,. .\o. ''*#MHHb\_ ._\ooob##HMMH#<' '"*#HM#b_ ,oMM##*"""`""` ``*MHb?,. ,dMP> "`*HMMH| .,9MH>dHHHc, ' ..odHM#P""'''"#M#_ .bHM#'' "#M#_ vMM` "HH?+d#HH#HH#Hb. JMP `*MM$:""""`"``' iMR `*M#\ MM- `*Mk. .____,_\,,,\o,\o,\ooo##HH: :MM `MH. v&MMMH#####*#***********"'"` :MH TMb `"` HM, :Mk `MH. |M} ?Mc HM| _ ?MH. MM .oMM' `HM\. dMP .bMM' `"HHb\. ,dMP oHMP' '"*HMH#\\. oMH' .dMM" ```##MM#bb\\.\dMM" ?dM&' `"'"####P' ,HM?' dM| ` dMP' dMP HM? dMP .Mk `` HMT .MM ,MR MM {M| MM. TM| &M| `Mk |Mk `4ML :MH '' ?R' vd#HH#b\. ,MM?'"`'*HM#o. .._,_,_.. HM! `?&M#\ ,dHMH#####HMHHo_ TMH#H#HHHMHH&MMMHbHMM$_:.... ``?*M#\ ":"""'"`'"`:ZMMMMMH*##HHMMHMHd\\ '*MH, .obMMMMMH**Mk. ''""#MH&\:MM, ,H&H"`dMP' &Mb "#MHMM| ,dM*' .HM? 9M? ``"' vMH^ ,MM'' `9M? .MM' ,MH' `9M? .__. :MH. ,MM' `HM: .HMMMMM#\ |MbbMH' `HM, |M[ HM} "*#"- TML `HMddHM*` .MM :?MMMT MM. .oddHMMMMM, MM. *&"'":MM`MM. MM. {M} `MM 9M: MR- ' {M| ,MMM: :MMM: dMHHM: ?dMMMHb\. ._\bdMMMMHb_ ,HM"`|MH. ```''``*#MHH#HHMHH**""' '`*HM#o____....... .___. "?' ?H: ._ ``"""`- "`*###HHMHHMHMHb,,,,,,oobHdMMHH#HMHdoooooooooooobdo\oddHHMMMMMMHMMH#dddMMM? ``***#****"`"''` `"""P"P"P"P"P""""***"'"`- '` '` '":""` ''I have created the above also with "the prompt"
Now remix with ai /someone :).
...now that takes me back
I could go further back, unfortunately I no longer have the original to show because there was just the one copy on paper which sadly did not make it over the years and a couple moves. But once I "drew" a cat -- or was it a bunny? So, so long ago, must be 30 years at least -- using my old Brother typewriter. Manual, not electric. And I am a two-finger typist. Still pretty fast now even with my arthritis and assuming it's a good day, but back then I had much less skill and practise. Took me a couple days to finish and my fingers cursed me all the way. Which is probably why there was just the one copy -- well, that and a lack of carbon paper. That machine was a brute. :sigh: Wish I still had it, I must be crazy!
Forgot to add: The main reason it took so long to type was that the animal in the scene was not made up of letters and symbols but white spaces where the typing wasn't. And getting all that properly aligned on a manual machine, especially by a half blind, self taught typist, was an absolute nightmare. No wonder I never tried it again.
It took me around 15 minutes to make this ASCII image, including drawing in The Gimp
and then processing this drawing. All done in Linux.
More ASCII graphics in my thread:
https://www.daz3d.com/forums/discussion/612576/artini-s-experiments#latest
I wonder if AI could create any art in such style.
...nice.
Thanks. It brings such nice memories.
Ai has big issues with hands...always does wierd things with fingers.
Though I'm seeing a real problem with the continued errosion in art quality and value as time goes on.
When people give up on being creative because machines are better at it... there will be nothing left for us to do...
Maybe it's already too late.
JD
If you are going to remix, bigger is always better. The more pixels you give it to run, will result in a much better image. If you make a portrait headshot of a person, they will look much better than if that same person is in a full body shot where their head now takes up a small portion of the image. You need pixels. One thing you can do is zoom in and crop the head of the subject, and run AI on this cropped shot. Then take the best new result and resize it and blend it into your full image. This way you can get significantly better faces.
Otherwise you might get aliens or mangled things. You can do this with several aspects of the image, including hands.
In this respect, there is still a place for Daz or other things. It is the only way I see AI not killing everything in the future. Because if you want a specific pose, you really cannot get this from a text prompt. At least not yet. Daz can give people the pose they want.
However, looking at the pushback against AI, I don't think AI will be taking over as much as some people expect. Not by a long shot. It is going to get regulated hard, and in many cases it doesn't even need new laws passed. The courts can already litigate AI art with existing laws and precedents.
Copyright laws don't get changed every time there is new administration. If anything, copyright law is one of the few areas that Republicans and Democrats will agree on without fighting too much, LOL. I just point that out if anybody thinks that a simple change of administration will open the AI floodgates. It wont. The laws are already in place to regulate. Only humans can hold a copyright, which is why AI cannot be copyrighted. Just to be clear...AI is not a human. AI does not "learn" anything. You might be able to copyright a text prompt you made up, but you cannot copyright the result. Simple as that. So pure text to image generation will never be copyrightable under current law. You have a slightly better shot at copyrighting image to image generation, where the original image is provably yours.
video
Back on Topic
@outrider42
Question will be, how easy it is to get something through for "made by a human". How many pixels do i have to change? (...) Grey zones may be reached quickly, and perhaps in bulk ("hand picked filter").
Concerning taking over - no. For a real thing, way more specific tools need to be combined, or the thing needs to be 100% accurate on any text input, in which case the general/generative property is in question. Perhaps there is something like the Heisenberg uncertainty principle in (sub-general) ai.
(Uncertainty principles are real considerations: https://www.sciencedirect.com/science/article/pii/S1566253521001081 )
The other edge probably is "the application", e.g. if something like this, possibly in an improved form, gets extremely popular, perhaps through integration with social media or other applications, with more tight integration into uis and more "you wish" type of interfacing, e.g. for finetuning. It will certainly generate some impact, with potentially imaginable effects like people buying less 2D and 3D art or mockups from others, also less automatic revenue from platforms with automatic license payments, and even without copyrightability, people might just "consume this", like drinks, and spread the results everywhere, which also might have some impact. Since training the system is and likely will remain expensive, it seems difficult to imagine an ad-driven cloud service just for the generator, but integrated into social networks, and/or if it fits the coffee bill anyway, who knows?
(For the euphoric ones: https://www.nytimes.com/interactive/2022/12/26/magazine/yejin-choi-interview.html)
(Careful with publishing date, Chat-GPT may have happened since, and this and that, but in essence we're not somewhere else yet, as far as i can judge it.)
The grey zones are a big question. It's like the banana importeur painting the truck green to avoid all the regulation. Well, maybe not exactly that, but i think people busying with computer graphics understand the direction of grey zones. Lawmakers might want a framework for data use, to avoid monopolism by design, and probably might be urged at some point, to prohibit mass exploitation of humans and their doing. The latter would also touch tracking in the internet in general (think ads, meaning whatever the data can be used for), and certainly exploitation for ai training and other. It also has to be seen in the context of proposals to open large networks towards smaller competitors, data/access-wise. Some speak of a potential separation of the internet into 3 zones, e.g. US, EU, China, at least regulation-wise.
Training data question on cc: https://creativecommons.org/2021/03/04/should-cc-licensed-content-be-used-to-train-ai-it-depends/
(Also contains some discussion of copyright for the output.)
These questions have already been answered by precedents in law. It isn't a question for science.
A court ruled that selfies taken by monkeys could not be copyrighted, and specifically because copyright can only be granted to humans.
https://cyber.harvard.edu/events/2018/luncheon/01/monkeyselfie
Think about what they considered here. Even if the photographer trained the monkeys to do this, and gave them the camera, it doesn't count. Just like training an AI.
This is further enhanced by the Copyright Office working to revoke a copyright they initially granted to a comic book with AI art in it.
https://aibusiness.com/ml/ai-generated-comic-book-loses-copyright-protection
This comic made huge headlines when it was awarded a copyight. But now it turns out this was never intended to happen. The Copyright Office stated that granting a copyright to this book was a mistake, and they make it very clear that AI art cannot be copyrighted because, just like monkeys, AI is not human. There is no question over how many pixels have been moved. If AI generated the image, it is not copyrightable, simple as that. Again, as I said earlier, you may be able to copyright the prompt of text you wrote to generate the image, but not image itself.
So we already have cases that set the precedent for law moving forward. There may be a grayish area involved with image to image generation, where you take an existing image and use AI on that. However, at this time, just going by the rulings so far (as well as public discussion that may influence these decisions,) even those may face copyright trouble.
I will point out that the question of how transformative a work is has always been an issue with copyright. You can look at the various lawsuits in the music industry as examples. Sometimes they win, sometimes they lose. But all of them involve humans suing other humans. I would expect courts to be much more strict applying these laws to AI cases, given what have seen already.
This is the most important point, and the point that bears repeating. AI is not human. AI does not have the same rights, in fact AI has no rights of its own. Any argument that tries to say that people can look at pictures and use that to inspire their style is missing this key point. AI does not get this right, so such an argument is completely invalid.
And that is before we even address the question of how the AI databases were created. That is an entirely separate problem of its own.
So when you add all of this up, AI actually has quite a few obstacles to overcome, including the court of public opinion. These obstacles may grow if law makers take steps to pass more regulations. Numerous countries and unions are already looking at making regulations as we speak.
someone on Facebook said they hate art thieves and posted angry emojis on a post where I said I "stole" some ai art and put the images through ai to make new ones
which looked utterly different to the source images BTW
I never even posted the originals or said where I rightclicked and grabbed them from
but
Stability themselves say the images are public domain so I really didn't steal anything
(they were other's Stable Diffusion results publicly shared on Lexica BTW)
it was just a cheeky comment
I sold nothing
I deprived nobody of anything
no crime was committed
I just asked who is the victim
This is from Stability's FAQ page:
"Can artists opt-in or opt-out to include their work in the training data?
There was no opt-in or opt-out for the LAION 5b model data. It is intended to be a general representation of the language-image connection of the Internet.
In the future, for other models, we are building an opt-in and opt-out system for artists and others that services can use in partnership with leading organisations. This model learns from principles, so the outputs are not direct replicas of any single piece."
Thus they explicitly verify that their training data contains images used without permission. Plus Stability is based off of Stable Diffusion, which also documented this fact. LAION 5b was trained on 5.8 billion images without any consent or knowledge of the artists. 5 billion is what the "5b" stands for in case anybody was wondering. This is the exact same model that Stable Diffusion uses, and a lot of the AI generators are based on this.
And that is another difference between a human and AI. No human can realistically consume 5.8 billions images and then replicate the style of every single artist those images come from.
Besides that, local laws will always supersede any EULA or terms from a contract. Like in the EU, regardless of what global return policy a company has, you can return almost any item because of their consumer laws. So Stability can say anything they want about it being public domain. Local law can override this, and they have admitted their training data contains unauthorized images.
Keep in mind that Stability AI has a paid subscription fee. So while you may not monetize the images you create, Stability AI absolutely does monetize their platform. This fact could very well cause them trouble soon.
This, all of it.
I don't care what AI users tell themselves to justify this and feel better about what they're doing, it's wrong and there's no way to make it right other than scrape the whole current database and start over. I found a lot of my pre-2019 Daz renders in the LAION-5B database searching "HaveIBeenTrained" by searching for nothing but Daz, Daz3d and DazStudio and then following the trail of images similar to results (never entered my name or uploaded an image), it didn't take long before my own images showed up, and all of them with my logo and copyright info on them. From what i can tell by quoted links most of them were lifted off Deviantart/Wix. It only makes me not want to create and post online anymore. Who are the victims? Every single artist who got overrun by the influx of AI spam online, every single person whose images are in the databse. We can't be so warped out of sense of ethics and morals that even consent means nothing in 2022/2023 anymore.
Of course. I am not here to contradict that. I just try to point at some of the potential, where things can go wrong, depending on which course lawmaking takes. EU might rule slightly differently, but there similar law is getting discussed.
It's for most the distinguishability question, with the potential of effective grey-zones, and perhaps the potential for lobbying in that context, which may render some of the current and future regulations ineffective.
Always with the distinguishability from copyrightable cases in mind:
- You get copyright for an image made by you, with ai somehow applied on? (Broad range of impact, subtle effects to complete overhaul,...)
- It's not possible or very hard to distinguish output from ai, modified by a human, from human-made?
- It's not possible or very hard to distinguish output from ai, modified by another ai or system, from human-made? (Worst case, perhaps, but also harmless intentions are possible, thinking of art programmed on a computer ~ depends on the program?)
Then there is side topics, which emerge, independent of copyrightability for a certain case, again not contradicting your statements:
- With "totally certain ai-detection" some artists won't be able to claim copyright on their (bread and butter?) artwork anymore, e.g. if the detection is done by a machine-learning system. Lobbying could in theory lead to exclusion filters, even if only on big platforms, effectively preventing any license-payments for your art, if you happen to be near-ai-art with your style. Of course that will fall into the category of people suing corporations or other people, but it may already take effect before results are there, if people can afford results at all.
- (Even not copyrightable images, if ever so popular and widespread, may change the market. That's not special, in terms of copyright, of course.)
The potential consequences might range from de facto copyright on images created in bulk automatically, to more or less moderate defensive copyright spraying, and copyright prevention for one extreme case, all to be seen in the context of automatic filtering, automatic license payments. Further perhaps, abusive legal action or intimidation purposes. Detection of modification and modified content, as well as the origin for ai-generated images, would be important, but it might be too hard to do (for the individual). Of course abuse will be at the risk of legal persecution, which may then depend on profitability or the intention to harm individuals, companies, platforms.
In that context the ability to go to court may be confined by distinguishability, and having the big players on the other side, with their potantial influence on lawmaking, while the effects might become severe for some people relatively short-term. Again no contradiction, just the question, if the law is effective in protecting, and who or what will be protected.
So if you will, i said nothing! Remains to be seen...
Here is a new way to make your Daz Characters speak with AI.
Which is why scraping needs regulation. For ai-training-data, some regulation is in the making, at least in the EU. Distinctions might be made for commercial and scientific use, though.
Especially "random images from the internet" would cry for regulation (or for court), otherwise artists have the choice to be exploited or not to publish, though if the EU decides, that it's necessary for the common good, people don't have much of a choice, but to found gated communities, where scraping somehow might be excluded by TOS, behind logins of course, with further costs, unless you build an image-search engine extra for "behind gates" (uh, oh, ... estimates). It will concern photography as well, not just artistic. While right now, some licenses don't account for use by ai, it neither means that there can't be made a case, nor that the next version of a license may be valid in excluding any scraping (other than searchability, if desired), thogh here law could still be made to overrule it.
(The same question does go for text, in the end.)
Looking up abstracts for "The Right to Process Data for Machine Learning Purposes in the EU", seem not to clearly answer these question. Discussions about "EU AI Act" didn't seem to result in a too distinct direction, concerning these questions. Of course there are some statements by some officials, concerning copyrightability, but they also claimed, that there would not be upload-filters with the EU-Copyright reform. It could go almost anywhere (effectively).
I'm curious what people think about AI usage in the DAZ eco system. I've seen some AI generated artwork that is quite impressive. But it saddens me that so many people will not really understand the difference between how we use DAZ and postwork when an AI can generate similar work in a fraction of the time.
A lot of people know what Daz Studio is and what it is made for. Main problem with posting Daz renders in 3D groups ever was those who wouldn't say it was made with Daz but tried passing it off as their Zbrush sculpt or otherwise not a premade model. Abundance of badly made adult art hasn't helped the case, but using and rendering premade models has its own place in the 3D ecosphere, whether it's Kitbash or Daz or realy any other script that allows mass instancing and alike, as long as you're honest about the process and know how to get an appealing result that stands out in the crowd.
Personally it really saddens me that Daz community resorts to "nobody likes us because we use Daz anyhow" as an excuse to go even further in not putting any personal effort into the result (sorry, typing words is not an effort, I don't call google searching a professional skill either). There's a lot that can be done with Daz Studio alone and a lot that can be learned on top of it without doing this.
@valzheimer - Yeah. I've been using DAZ as a hobby for many years and recently storyboarded an entire film project with it. Many people were surprised and impressed with my storyboards and have never heard of DAZ. But I was really depressed in seeing one of my colleagues and my own brother use AI to create artwork for their own projects in record time. I love using DAZ but it made feel like I''ve wasted so much time and money all these years if people can just type some keywords into a text box and spit out great looking pre-vis artwork.
TBH ignore them all, while you can. I'm not officially an artist in anything (lies, in a way, as some things i did are considered art in general, but not for a living, and not in this context yet).
Recap of that type of stance: "Any 3d-artist who don't write their own operating system is void, forget about OS, design your own chips and hardware, or i won't even bother to consider you a valid entity in this universe."
To hell with that Dalek attitude.
Even further, i reserve myself the right to do as cheesy as i dare, renders, computer games, anything...
Do they do storyboards out of nothing?
(Edited for better readability.)
Text-to-image resembles "do what i want" (yes, not yet fully), wheras using DAZ assets means "exploring other's works, in order to create a new one".
Of course artists will always seek a way to create something, regardless ai. I'd still bet that the predominant application will be more like "do what i want", with the tweak of "i want what it outputs", potentially ending in disaster. This could become a great reprogramming exercise.
In the example of my colleague, she used Midjourney to create the artwork to accompany a 23 minute podcast. I was definitely impressed with the look and presentation which is why I started feeling depressed about how long I spent on my own storyboards for my film project. At some point will this AI make DAZ obsolete? Will comic artists be a thing of the past? Animators? Where will it end?
When will it get to the point where it starts creating false situations amongst real people