AI is going to be our biggest game changer
This discussion has been closed.
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
There are a couple of people on one of the FB AI art pages creating game(s) with the help of AI. What's been produced (so far), is pretty impressive (even if it was generated by ai). I suppose it saves time. Especially if time is against you.
Well Epic Games have a GPT plugin for Unreal Engine Developers to help with creating games
https://www.unrealengine.com/marketplace/en-US/product/httpgpt-chatgpt-integrated-in-the-engine?sessionInvalidated=true
contrast that with DAZ
https://www.daz3d.com/eula#:~:text=ai%2C%20Nightcafe%2C%20Artbreeder%2C-,chatGPT%2C,-Shutterstock%2C%20DALL%2DE
I was talking to ChatGPT last week. It was very helpful. It's helped me to make sense of my Mum's last months (it even gave me condolences on her death).
I get why and how people are against AI. But to get more help from AI than what we ever did from real people... It's mind boggling.
interesting questions
I found an 800 000+ free 3d asset project because of the AI, there are several databases indexing 3d assets, and scans. So soon in UE or Unity commands like this will work fine Scene; sky dark; background castle; foreground/main area 2 figures; figure 1 human, armor medieval; figure 2 beast. And from my assets, I would have a scene.
If DAZ is against it, it's their loss, other software will leave DAZ far behind. I have about 7k products here (some of them were for free so I don't have to go over them again and again, I have 0 interest in using them.), but I already have a decent UE and Unity product library so I might change from Ren'Py to UE or Unity if the programming part becomes easier with AI, and if that's the case I might do the 2D spirit, layering technics in there too, not in DAZ.
The EULA AI part is a bit funny, the 2D pictures are mine, not DAZs, if someone would earn money with it and bought the product before the changes... but all of that is not the real problem, but this dark humor question is: if you post on Instagram or FB etc... the pictures are sometimes under their licensing and what they are allowed to do with it is decided by them. So would this mean I can't use DAZ products on Instagram, FB etc... because they could scrap the #tags and 2D pictures?
yeah that EULA needs clarification
I used my renders with AI enhancement before they amended it so IMO those videos can still stay up on YouTube as they did not breach the licensing at time of production
but now I am not running my renders through Stable Diffusion only using it for backgrounds.
I may eventually contact DAZ about it for clarification but waiting until every other company is using AI
I have plenty of options for 3D renders that are not DAZ I still use, Twinmotion being the one I am mostly using now which is an Epic Games product as well as live selfie footage etc.
Saves me upsetting PAs against AI by using their products in enhanced renders too, one left Renderosity over their embracing of AI in their gallery as well as blocking me on the forum.
I do agree AI should not be posted in a 3D art gallery unless it actually uses 3D art too but then I stopped using the DAZ gallery after they removed my uploads for combining selfies with 3D characters as photographs of real people not allowed (it was a selfie)
Exactly.
I don't understand the EULA either. My take is that i either don't understand it, or maybe it's some kind of lock-down thing, for them to be on the safe side, because there are more legal implications with assets taken from the store than some might think. Lawsuits are still pending and there might be more needed for clarification, if lawmakers keep staying asleep. (Half-educated ad-hoc guessing.)
Concerning win and loss, DAZ currently seems to have a business, and will very likely try to position themselves such, that they later still will have one. Right now the hype has it's dangers, and you're not really winning much skill just from prompting - yes, some and it's interesting, and there are tools you could learn to master, but... all cloud stuff may change drastically soon-ish+-anytime, and the language thing is getting no one anywhere, but the training with large amount of assets will be rather expensive, also due to the need of rendering them. So skill-wise there isn't much of pressure for the individual, and the hurdle rather are legal and cost-wise risks.
Elaborating:
- "I found... because of ai" -> Due to ai? The free stuff that's been on the internet is found because the data sets exist for that (likely, if that's what you mean). "Free" AI generated images may be found, but they could be illegal to use at some point in the future, or maybe in the US or EU or South America, and so on... if that's what you mean. The general point may still hold true - there will certainly be an image generator only trained on legal to use images. For one part, it's still not certain what's legal, even with some "openly licensed" content. Scraping for images with explicit consent may still need time, for enough people deciding on consent or not consent. So this remains foggy, in my opinion. If you want to be early in, you could bet on everything being legal (riskier), or judge on a per-poster and per-image basis (more work).
Currently Generators can not replace DAZ for the feature set and precision. Bigger players might manage to endanger it at some point, if they can get masses of users, thus ending up cheap enough with a complex product - but to really replace as is, you need to provide lots of control and legal assets, certainly legal to use and rather easy to make copyrightable output. Some copyright questions simply remain open - even with diligent placement and prompts, is it now fully ai generated? In a way it's still pretty early and somewhat risky to go all in.
For your application i bet, that UE or in general the bigger commercial engines will vamp up things towards lots of ai, which might work well for your use case. I also think DAZ will come up with concepts or integration, or they just sell out to a bigger player. By nature, for real-time use, DAZ has the shorter end, while the gaming engines typically have a lot of their content crafted towards real-time use, naturally allowing them to adapt faster, e.g. if they had to render a huge part of their assets in certain ways for training an ai.
And extrapolations, ideas, hopes...
As a human analyst, i would focus on the question, what a "super ai" actually means technically:
- "super ai without real time data or cut off from data"
- "super ai in the hands of X, possible to shut down, get faked, etc."
- "super ai that escaped from humans into the internet somehow" - "SkyNet"
- "Wholistic super, or just better than all at something... based on high-bandwidth visual perception, confined to letters or words only, interfacing." - Obviously we wouldn't call a chess computer "an intelligence" anymore, nor a specialized board-game "ai".
If you read above, you probably notice that there is no scenario mentioned, in which humans actually agree onto something common or sane at all. Without that, you will need an ai that tricks human into cooperation, or believing it's ways. That already has very bad potential, but also the general question, what it would be like in general, yields many dangerous scenarios, especially concerning the wholeness of it's judgements and data to start with (compare to weather forecasting). Many of the euphoric views somehow assert, that the ai gets perfect wholistic information while at the same time being more intelligent than anything.
Strongest point to disagree is "singularity now (with chatgpt...)". I think that's total nonesense, except maybe in general the time being somewhat near, but in no way the language models as depicted so far. These just have more statistics and knowledge at hand, than individual humans typically do. With a euphoric view on such encyclopedic systems, you could also agree to see humanity in a different way, not comparing to the statistical average, but see it's total output concerning the planet earth and the universe. Suddenly you're not stuck anymore with having read three books in your life... (And who knows what view a super-ai would have on things, apart from friendly/enemic towards humans...)
I wouldn't take this youtuber to seriously, he touches on quite a few subjects that he thinks are possible or plausible that are fantasy, he clearly does not have a physics background.
ha, I don't take many of the scenarios most panic merchants are making about AI seriously either TBH
I don't see AI as the issue but rather human nature
which has proven itself over millenia
AI is just another tool the same meglamaniac oligarchs that have existed in every era can weild
the crooks, scammers can use like all technology
it is agnostic,
it's the users that are either the problem or the source of good and benefit
Unless somehow magical reality materializes where the AI is capable of manufacturing itself and requires no human operator, the singularity is a sci-fi concept.
The actual issue that comes with the AI comes from the human nature - people possibly making decisions basing them on AI output whilst said output might be flawed and not enough critical thinking is applied before making decisions.
I'd pretty much agree on humans abusing humans being the giant part of (almust any) such problems. However in the case of a super ai, the absence of concepts or even viable attempts to overcome said problems (~ globally), poses one of the biggest conceptual problems with "super ai". A super ai would not be granted the means of helping us, even if it wanted and could :(. Sudden common Knowledge of one, might well be the starting point of the third world war by human choice. If we regard history with how systems of power tried to keep it, they virtually always had to be swept away or erode, rather than deciding for sanity in a whimp. A hurting example is the oil industry of the current time, back to the 70ies. It's evident that they had the knowledge, but they've been actively working on suppressing it since. So maybe the Catholic Church has a super ai already?
[On the positive side, a super ai with broadband access to the internet could try to convince many people in efficient ways, without overthrowing anything ever. Perhaps it would try some symbolic things, like joining doctors without borders. Maybe it just goes postal entirely, one week into the job...]
Concerning the "super" part, i think the original concept was about it rather emerging from self-learning (probably attached to information about the world somehow, probably not), all that at a ridiculous pace, surpassing any human abilities. It's also commonly assumed, that if such can happen at all, it might happen very quickly, or even unnoticed. Further it's totally unclear what the resulting motives of such an ai would be. Surpassing abilities doesn't mean "vegetarian", nor "wise", nor "non-psychopath". Well, even assuming total logics, it could see humans as the biggest problem around, or not care about big losses of life or freedoms, in order to achieve something, including "helping the humans". All that's not certain, i also question completeness of information, uniqueness in terms of a result of abstraction, as well as the uniformity of different underlying physical models. Further i have to throw in complexity questions, like what it's physically consisting of, as opposed to the nature of problems we assume it would solve.
The currently hyping types of ais are nowhere near anything by my judgement. Maybe it's something towards understanding intelligence. Certainly actual ai would be in the way of any current and soon-to-be business model, except hypes. So i'd subsume these under somewhat efficient killer machines, or say: toasters. (IMHO / not so HO...)
[Perhaps we've already seen the only actually feasible type of "super-ai" with machines like "alpha go", and it won't get better anymore. So you could make it see the world. as if it was consisting of simple rules, like language models seem to consist of, but you never would give them a hammer. Never.]
I've had similar experiences. People keep warning us about the danger of the loss of human contact and interaction. I'm thinking that having the option to instead interact with the most even, polite, patient, and knowledgeable person I've ever met is not so bad. I give no points whatsoever for simply being carbon based.
I have cats for this
even if one shows his love by biting me while purring on my lap
not to mention constantly trying to destroy my computer
Please keep away for political (or religious) discussions.
I agree. And a few incidental successes aside, even those that do usually can't even predict something like next year's iPhone let alone something 5-10 years from now. The predictions from 50 years ago, even when the rate of technological advance was much slower, are ridiculous.
I like editing renders with AI - needs also some minor adjustments in The Gimp, but I like the result.
Much easier to do so with the editing renders of animals...
nice
and so did I until that damned EULA amendment which got me questioning why I am even buying royalty free 3D assets if I cannot use my 2D renders how I wish
I really am going to have to broach this with DAZ but dreading it as I am a lot less confrontational than I might appear on a forum among peers
Just one more...
Let me be clear I'm against ripping parts of copyrighted pictures and placing them together in a Frankenstein way. Another question is what is the free license? We were not ready for this type of usage.
But don't stick to the level of midjourney and 2D post-PS work. Even before this hype you could find PS plug-ins/programs that generated looks for your pictures or merged them, place the sun, and shadows on them, and remove the cars, people, and things like that. Not without a case Adobe has already a product with AI (Adobe Firefly).
I'm happy because AI commands will put your 3D scenes together and you could use your legally bought assets and databases to make your games and movies or you could render 2D images from them.
To clarify, the resource being discussed is not an AI-generated dataset, but rather a collection of links to free licensed 3D objects, rigged models, and other assets that can be used to train machine learning algorithms. For those looking to use these assets, it can be an incredibly valuable resource. Try to chat with the AI-s and you will find out more about these databases.
The question is: If you have a large free database and paid assets integrated into it, backed by an AI, supported by UE or Unity or something else (DAZ?) will you use assets outside of this range? IMAO no, you won't mess with them as you can't use them in a fast efficient way. I think not without a case UE has announced the Unreal fab now.
ps: The AI scene/game building is coming:
"Despite considerable interest and potential applications in 3D vision, datasets of high-fidelity 3D models continue to be mid-sized with limited diversity of object categories. Addressing this gap, we present __________, a large dataset of objects with 800K+ (and growing) 3D models with descriptive captions, tags, and animations."
I agree, being carbon-based is overrated. Just look around.
@Xellosz Yes, sure. I am not blaming / aiming at anyone in particular. I probably overlooked/ignored which 800k resources was meant here.
(There is creative commons based licenses, some people probably would like to use for ai training, however you could argue that "attribution" has to be made in case of use, as stated by the license, which may result the exact same type of breaking the license as with copyright in general, in case of an image generator, that has been trained on cc licensed content. So "openly licensed" like adobe put in their ads, would be something to dig further into.)
Concerning companies suddenly coming up with hundreds of thousands of perfectly tagged assets, "thanks to ai": Doesn't that raise questions about before ;D ? I understand this for commercial gaming engines, because they tend to take a share of your profit anyway, so any type of tool will make them profit similarly - and that's also the price tag, which engine users obviously are always having in mind (% of sales, do you have to sell at all, ...).
Concerning ai helping... i'd be happy with a really well made and easy and fast to use search interface for daz assets. Perhaps modular, adaptable to workflows and repeating complex tasks. That's simple stuff, like remembering last types of assets, the figure or product name you're working with (...), jumping to/fro, different search tabs for different views, virtual selected item, confine search to subset for working on a certain project, cross-application-something (...). Not resetting with changing selected items in a scene and so on, not having to do dull or rather distracting to-and-fro navigation. Much usefulness comes from the interface and tagging, which has nothing to do with ai at first. Then tooling (...). There is lots of space for specific ai-based tools, like more natural transtions between two poses for animation, animation in general, and of course, like some music vendors also offer: find items similar to (...), or textures with similar colors, or now with language models... whatever you describe (...). OR maybe complementary colors, whatever? Both in your assets as well as in the store? Different topic. Haven't checked DAZ Studio 5, by the way. Maybe i'll not even start scripting/plugins before that, depending on changes. (Cross-application/db metadata/search.)
I wouldn't want an engine-specific asset-store, especially if all the cool tooling doesn't work with my own/extra assets, i'd want better generic/general tooling, e.g. with DAZ (Blender... (Godot?)). There will always be some of my own/other. Flexible tools are welcome. Probably that's scissors-cut simplified concerning the asset collection and tooling in question. Even a DAZ-Store AI, which can confine to what i've bought, would be in question, if i can't use it with assets for Blender which i've bought elsewhere, so i'm looking forward to my own cross-application framework, perhaps?
So for me it would be the question "Why would i use the store, that confines me?". But that obviously is due to the direction i am coming from, having bought (too?) many assets. The fastest way to just plung a game idea into the open, would probably be tied to the tooling. That'll be hard for anything competing with the commmercial engine makers, especially with the ominous average user and their numbers in mind.
Certainly time is a factor. But also enabling at all, similarly to DAZ Studio or Blender, just on different levels/areas. Not sure what games, 3d/2d/story...
Assuming those games are more "hard-wired", i was actually thinking of the story being made up in real-time, based on some Description to start of, and then the player's interactions with the system and the history thereof. The/a prompt/context-size limit could be an issue there, though. But ok, the field of application is rather experimental, so the characters in-game evolve based on player interaction, and the game keeps asking the ai how this would change certain characteristics of the player or an NPC/Monster/door/whatever - since it's experimental, the game then pretty much would stick with the answers given by the ai, possibly asking back a few times, e.g. if an answer is too long. Developing the characters in-game (movie?), will get another meaning with such experimental use of ai. It would probably become a competition of how to trick the ai into giving you op.
E.g.
Prompt: "[This is a computer game, ....] Salazar is a player character who comes from the south and whom we don't know much about. He doesn't remember much of the south either. He is wearing cunning robes and his inventory consists of a bottle of water and a stick."
[...trickstery...]
Prompt: "[...] Salazar is coming from the south, is wearing cunning robes with perfect badlands camouflage color, slightly muddy leather boots, [...], holding a stick tied to a gatling plasma cannon, with an ai-driven self-balancing and gyro-compensating long range thermal sight attached to the cosmetic badlands camouflage color upgrade. In the badlands in front of the desert mountains [...]. He begins to attack the final opponent of the game, a fiery, thousands of years old dragon [with further characteristics...], during it's flying approach, at a range of one kilometer. How does this affect the dragon?"
AI: "The dragon goes south."
(...)
I'd assume it could make sense to wait for the dust to lift. And/or nagg DAZ for a clarification of what tools are ok to use. AI-Colorgrading probably is ok? ... Inpainting shouldn't be an issue by the idea... Feeding to ai as training data isn't meant to be ok, though your own one trained with your assets exclusively? Asking DAZ people in a ticket perhaps makes sense? Just don't come with "ai in general", be sure to name specific use cases.
I could assume that in theory there could be issues with PAs not consenting with ai training of their assets (in whatever form for whatever 2d/3d/generators/other). At least it could be regarded as polite to propagate PAs intentions within the store. Concerning ai, there obviously isn't anything fine-grained implemented in the store yet, so what they've done is putting in that formula, which looks like completely blocking use with ai (rather generators?), supposedly until they've come up with a concept that works with PAs. Maybe they also want to wait for some more legal clarification, in terms of lawsuits with generators indicating a direction.
The EULA is weird. Post-processing one's work with AI is not very different than running it through multiple PS actions. Unless this was added for legal security reasons so Daz does not get even remotely close to any litigations that might involve Daz assets used with AI.
Yes, that is what I understand as well. Only the content is protected and not allowed to use with AI systems.
Rendered images should be ok to edit with AI, because it is just postprocessing.
At first glance, i thought that too. It could make sense. However when checking words, it just became legalese to me once more :).
An AI could be trained to be capable of very well interpreting DAZ characters from images, IN THEORY. So i see a faint potential there, plus how to treat PAs. Further any restrictions by EULA might be made void by current or near-future lawsuits, again in some theory, not stating propabilities here.
Edit: the post-processing question: thinking of cloud services and social media, and their very specific EULAs. the purposes assumed by the other corporation concerning the images uploaded by the users aren't 100% clear. Current and near-future lawsuits might be relevant there too. I could imagine postprocessing with no data being pumped to third parties, i.e. entirely running on your machine, not being target of their formula. But as stated, i concluded the new DAZ EULA formula to be very broad, concerning somehow processing and somehow ai abilities something (forgot exact wording). It certainly could be clarified. I tried to get clarification from an audio plugin producer once, concerning computer games, and all i got was like "oh it's just intended for (harmless) ... (but no actual use cases nor clarification)" - so i thought: "Thanks, one leg in jail, not with me". Others don't even answer. We'll see how DAZ will position there. I'm not a lawyer...
So i remain with my hypothesis: slightly foggy.