AI is going to be our biggest game changer
This discussion has been closed.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Sure, yes, I agree. But let's talk again a year from now.
That looks like autotune sounds. There hasn't been a decent sounding autotune song since Cher's Believe.
"I'm sooo confused..., I'm beginning to grasp the answer to Life, The Universe, and Everything. I even have an inkling as to who and what I am. Aummm... But what I don't understand is who the heck are you and why are you trampling my roses?"
----
It has been noted that he had worked as a programmer way back in the days when the * was a return for an unknown quanity or something like that (details are vague that was back before Abe designed the first computer)... it's ascii (or mayb ebcdic) code is 42.
Talking about ramen...
Oh my god, how could I never have realized that?!
..and then there is this:
I'll try to be very short, but not too concise here:
1. AI learns. NO it doesn't. It's dead fish in the water. It's "PRE-TRAINED" like ChatGPT. Anything else is misinformation. The most you could say is "it has learned [from...] somehow".
2. Creativity / new. The parrot analogy is more correct. It's partly, based on prompt, creating "new stuff", but not really. It remains to be encyclopedic parrots. That's the tech part. Compare ChatGPT (and the likes) being a flat area in 3d, you grab it slightly and so on and it looks like a pyramid for a moment, but it swings back directly afterwards. But creativity is a torus, at least.
[Edit: joke with topology: one dimension higher, spinning in time, and if you touch it, it'll immediately attempt to disguise itself as a flat 3d-area...]
I won't be sorry for this statement, a couple of years from here, because what we now have is this, and if there will be something better, that even doesn't need to kill people to exist, so be it.
wolf359,
Thank you for sharing that!
I lost count of the number of times he said "storytelling" and "storyteller" in that video. How many of us will make the cut? For some, the gulf between the left side and the right side of the brain is vast. There's already so much navel-gazing, second-guessing, anxiety and faux-morality surrounding the word 'ARTIST', what more about being a ... (whisper) storyteller. Or maybe most people will just be happy creating still images. There's safety in the familiar.
I do give him credit for throwing down the gauntlet: in the face of changing technology, how do 3D artists reivent themselves? It's probably safe to say folks will be spendng less time 3D modeling in front of the computer and more behind a camera, out in the real world, taking videos. Come to think of it, that doesn't sound bad at all.
Cheers!
I find group chat videos very hard to listen to in general
but I will try to watch this one later or read the transcript to be fair and hear all angles on the subject
OK this stood out to me
you know behind the camera and I just feel like a lot of minorities in general are are fighting for our
20:58
place to be seen in the industry and now with gender technology like this we're starting to see you know um photography
21:05
uh fashion shoots um and portraiture of minorities that
21:11
aren't even there or aren't even real you know I've seen um fashion shoots of like African models
21:16
where there were no African models you know and you know I've seen also uh like recently a photographer was outed as
21:22
using AI imagery and like none of their photos were real but they had pictures of minorities in you know their settings
21:29
and this is dangerous because now we are viewing minorities through a uh a
21:36
non-minority lens and as we know all of these AI imagery are are drawing from data that's already
I am pretty certain they are referring to Shudu_Gram
a DAZ3D Genesis figure
Sorry for your loss.
Not sure how to share youtube Vids, but this explains the legal.
https://youtu.be/gv9cdTh8cUo
That's deep.
<iframe width="711" height="400" src="
title="GET PERFECT HANDS With MULTI-CONTROLNET & 3D BLENDER! This Is INSANE!" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen></iframe>
For stills in the next 2 weeks Midjourney V5 will be released and the realism is off the chain. Hands are finally beginning to look like real hands.
well I seem to have broken my local installation of Stable Diffusion
so might be going back to doing DAZ iray renders
(I was doing renders on my other PC but mostly Carrara and Filament)
Midjourney V5 sounds promising.
@WendyLuvsCatz I assume that if you don't have a GPU, Stable Diffusion won't install?
I'd love to run Stable Diffusion locally, but it's just not installing AT.ALL *weary groan*
yeah doubt you can run it locally
there are some CPU based builds but I don't know much about it and they would be awfully slow and resource heavy
am actually having issues myself right now with my install I am trying to fix
Here is my 1972 airport shot.....problem is she has something like a cellphone LOL Midjourney V4
for the fellas saying "AI/SD is just a bunch of images and do a collage of your prompt",
how you explain this variation of my Daz Render considering nobody has the same character, hair, environment, light setup, skin material as a full product?
if your conclusion is still right then who collected my local images to internet and train in another PC to craete a safetensor file and that way recreate my original render to an AI output?
If you know how to use them, DAZ and Stable Diffusion are a marriage made in heaven.
And where do faces or watermarks come from?
I just pressed ctrl-w by accident, so you get the very short version.
These generative systems clearly are encyclopedic, given that they expressly train to reproduce source images as close as possible. Due to specifics with tagging and the language model, and some synthesis/abstraction happening, it may not be possible to prompt-engineer every source image to display "as close as then possible", but for some elements and sometimes full images, it actually works. And this is the bigger points:
1) Training on images against the consent of the rights holders.
2) 1) for generating images competing with the images trained on. (Clarification: with a massively scaling computer system and commercial use for basic stakes.)
3) The bad cases for rights holders (and in my view: for society) are the crucial point, given the properties and abilities of these kinds of systems. You may always find a good case for you, but that says nothing about the whole thing.
I'm not judging "ai" itself nor people using "ai" in general. I'm not part of that "discussion".
can't remember if I quote you in first place about my sentence, because in the page 38 you are not the user who told that, but thanks for your pov
Because it was learning from large sample of images containing watermarks, therefore it "assumed" that having something like a watermark is legit part of the image.
It really isn't unlike fan artists learning by copying their favorite game or comic art. Without any deeper understanding how to draw they often don't realize that they are also copying shortcuts and often errors common in comic books.
Which is why art teachers almost always say to not do tracing because you can pick a lot of bad habits this way. Looks like the machine learning runs into exactly same problem.
It's both amusing and fascinating to see a machine making mistakes similar to what a human would do.
The damaged but still visible watermark on AI generated art is simply clear and verifiable evidence of the copyright infringement of the artists source material.
Well, if it lacks ability to evaluate things and repeat brainlessly as the machine does, it is. But it's happening because when it tries to copy style of an "artist X" unlike a human copyist it can't realize that the watermark is not a part of the style.
If you had a human copycat and somehow they could not realize that the artist sign is not part of the style they'd run into the same problem. Except of course a human can realize this.
Indeed, but even without watermarks popping up, the technique is "this" kind of thing. Not meaning blunt collage.
What i'm saying is just, that it's not just "like" or "looks as if" [...% of the time...], but that it's base is encyclopedic by nature, and that it also is "like copying", even if was like [something that humans have or do] at the same time. If you add filters to remove watermarks and correct hands, say you'd be cheating, because it's part of a system, but not of "the brains", the problem would still be the same. It's not just a legal fun question for me. If you then start adding special filters to ensure it looks different, you would have a different beast, in a way, because the system would then "have a style" (in a way), and that wouldn't go well with the "do everything [cheap]" approach. The kind of system is bound by it's nature in multiple ways. Maybe, maybe the next version will be real-time and true intelligence, and it'll just do what it wants, and no two prompts will ever result in the same thing [careful: can be reached without any "intelligence"]... assume, a system that doesn't do what it wants in a deliberate fashion, likely isn't using much of intelligence, isn't being creative, for simplistic criteria.
Even without that i would still opt for any "encyclopedic-like" "super capable" and potentially "super scaling" system, not to train on content without the explicit consent of rights holders. Damage to society will be a preset. Not because we couldn't have a society with "free training", but because the transition towards there is no fun game to have.
@Zilvergrafix Not unlikely! Thank you for the opportunity to throw in these points at random, then.
I haven't found any results that stood out on a search engine about running SD on a CPU.
HOWEVER, I now have a working version of SD on my PC! ^^ - Although I had to edit the executable file in notepad (I know it's not an .exe file, but it's the one you use to start it up). But, I've only got to try it once (I didn't get a good result either). But yeah, it is resource heavy (although I wonder if it's heavier than rendering a very detailed scene in Daz). Plus, I'm not a big fan of using cmd (although using it makes me feel more smart than I actually am when it comes to computers ^^' ).
I've now got a lot to catch up on (i.e. learning about models and ai terms).
I hope you can get yours up and running again as soon as possible :)
I cannot get my Visions of Chaos version of Automatic 1111 Webui to run
I managed once bypassing VOC and using the bat file instead but now that throws an error and typing what it says to do in CMD. throws an invalid syntax error
I have no knowledge of such codey things so just at a loss, I can still use the GUI version