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
And the "Stacked Generative Adversarial Networks", which generate images based on brief descriptions:
https://www.youtube.com/watch?v=rAbhypxs1qQ&t=108s
And while we're taking a moment to admire all the tools AI will be giving animators in the near future let us also take a moment to bask in the glory of this example of AI controlling all aspects of the creative process.
"I'm full of ordinary sounding words!"
Not feeling threatened yet..... "Exploding kidneys?"
Well if you want to start feeling a little bit threatened, there are many computer scientists who believe the gap between laughing at the silly monkey antics of AI and "oh crap! Singularity!" is terrifyingly short.
I thought it was pretty good but I don't understand why they didn't generate random voices for each character and the narrator as individuals.
Yeah, the rate of change in between is predicted to be exponential. "Hey, Kids! Singularity's a comin'!" Also, the "Leonardo" machine (in Italy) will leapfrog us to the 10 ExaFlop scale in one machine in about two years. Ten to the 19th power Floating Point Operations per Second......EEEEEK! This will surely test Heinlein's theory from The Moon is a Harsh Mistress! Would probably render pretty fast.... My hope is that the AI's will keep us around because we are fun to watch.... "Hey, RT-95847! Did you see what those little Monkey-People did TODAY? LOL!"
Oh, I don't know. Ever since "Mankind" could dream, or think about artificial beings, intelligence, etc, there have been predictions of the potential power of Artificial Intelligence, computing power, whatever. I remember back in the 1960s they said we'd be driving "flying cars, or driver-less cars," whatever. We're still behind the wheel of our own cars.
I bet that aren't the same as thought of AI "coming soon..." about 20 or years ago, right?
I sat in front of the TV when the first man walked on the moon.. I guess I sit in front of my TV when the first human walks around on Mars... but I don't expect to be around when real AI will come around.
I don't think we will see anyone on Mars, the way we are going Earth might not even be populated too far in the future
I'd love the idea of being able to hand off some of the more manual or technical bits of the creation process to AI. I know the argument about "taking the fun out of it", but my fun, and the part I enjoy, is world building and storytelling.
I only use DAZ Studio to help me realize this as a comic, but if a robot wanted to pick up the slack I wouldn't complain.
AI is fine for generic stuff like denoising and sharpening, or generating random face morphs or landscapes, but it's still a long way from reading our minds.
There's a lot of research being done on AI safety. Nothing will likely blow up using AI to make images, but the same issues of getting users to understand every single possible outcome of their instructions apply. AI requires a reward function, so that it knows what outcomes are desirable and which ones are not, and it will follow the most efficient path toward accomplishing the most desirable outcome. If you don't set your goals very carefully, you won't get what you thought you wanted, and users need the language skills, whether they be programming language, or eventually natural language, to be able to provide instructions. The language people use every day is the language we will expect machines to be able to interpret, and AI is a long way from being able to keep up with changing language patterns. If it understands what a millenial asks, it will surely misinterpret what a baby boomer wants.
If anyone's interested in the topic, there are a lot of videos on AI on YouTube in the Numberphile and, especially, Robert Miles' channels.
Sooo... a bit like Djinns manage to get their "master" give what they ask for, not that what the really want to have. Example for that is a X-Files episode.
In the end the danger with AI won't come from the AI being malevolent, but humans just to dumb and/or greedy to ask the AI to do the right things
When I worked in IT a lot of people were saying wouldn't it be great if we didn't have to write code, just tell the computer what we want it to do. But I think one of the problems with this is how careful you have to be do give precise enough instructions in natural language. If a computer could understand english as well as a person it would be able to misinterpret intructions and do the wrong thing just like a person. After reading the specification for a system we often had to get back to the users to ask, what about that situation? what do we do when this happens? etc. This is why we have long complicated written laws and those small print user agreements that most people never read, so everything is spelt out in detail if there is a dispute. I suppose the legal system is a bit like coding but for people instead of computers.
There was a thing a few years ago, a large collaboration of universities and tech concerns IIRC, and they programmed a setup of some sort to "paint" like a well-known artist... I forget who, might have been Van Gogh.
Part of the experiment involved 3-D modelling and printing - the "paint" apparently had carefully constructed depth patterns that were tied to the "brush strokes".
I haven't spent a lot of time in auto assembly plants but I have watched one of those embroidery machines stitch a design on a baseball cap. It is really wicked, several needles moving in and out rapidly. Still requires a designer (or at least a "typesetter") to produce designs that the machine can follow.
My watershed moment for portraiture/photography came about 30 years ago when I passed by a small photo studio on a busy street. There was a large 4 ft. x 6 ft. say, possibly larger "corporate portrait" sort of picture in the window, along the lines of the ones in Washington D.C. - past presidents done in oil, that sort of thing. Except it was a mounted photograph from (probably) a 2.25 or 4x5 inch color negative and it was printed on textured paper.
I love large format photography, but unlike classical portraiture this large color picture was absolutely lifeless.
I like the way current computers trace out light and shade... it's often a surprise and in a good way too. I don't mind taking a hint or tip from the machine and from there there's nothing stopping me from adding my own, painterly twists and turns/emphasis etc.
The AI doesnt have to be creative itself. It just has to have good examples to use. If I tell the computer to make Inaru lie facedown with her hands behind her head, the AI needs to have images of people described the same way, and pose Inaru the same way as the person in one of the images. If it’s really good, it can do the facial expression too.
If I tell it to make Inaru wear a business suit, it can find something in the library that is classified as a G8 outfit and business apparel if, you know, the metadata was actually well maintained.
If I tell it to make Inaru have short brown hair, it can find a hair figure that is classified as short hair, and search through its material presets for one that has brown in the filename. If I tell the AI Inaru’s hair is long then it can update the metadata to say long hair instead of short hair.
The AI won’t be good at doing your thinking for you. It will be good at taking what has already been done well, and doing it again for you.
If it's that simple, then the artist becomes unnecessary, and viewers can create their own images using those canned inputs depending on what they'd like to look at.
Well these new Disney toons to Disney real movies are pretty close to exactly that. I've not seen them and won't collect them like I did the classic Disney toons.
I take back what I said earlier about AI...just tried the new AI filters in Photoshop 2021 and can't wait for more. Amazing stuff.
It looks like it has the same feature to colorize black and white images that was in Photoshop Elements. Could you tell me if that works on people with brown or dark brown skin? All of their example videos had people with the exact same complexion and made me wonder if it makes everybody that color.
I've only tried the new Neural Filters like the facial expressions and the style changes so far, but I wouldn't be surprised if it doesn't color people with darker skin correctly. AI have had trouble in the past with separating darker skin from backgrounds and such, but these improvements are a good sign. Maybe someday you'll be able to tell it what race the person in a photo is supposed to be and it will give you dials for specific adjustments.
Yeah, I don't see how anything could go wrong with that.
FaceGen Artist Pro does a decent job of that already with both the estimate base off the photos you supply and FaceGen has morph silders to alter the face geometry you supplied based on ethnicity data sets it's has in it's program data set. It's all based on real world geometric data and nothing to do with the author's or anyone else's personal opinion. It also has a feature to generate generic random faces based on those ethnicity data sets or all of them together as a world set.
here a awesome video about cyberpunk 2077, all the "face animation was made by AI
wow it's a big change and the characters really looks amazing even the hair look very cool.
Damn, that is amazing
The best thing would be when it becomes like StarTrek.
So we can walk into a room of Genesis 12 figures that are 3D for real
Walk among Giantesses...
yeah, the game industry is really pushing hard the 3d and programing world, the more "complex" the games the more complex programs you will need to keep up with the demanding.
'AI is going to be our biggest game changer'
People have been saying that for over 40 years. If it ever gets here it will have its place although I doubt it be in the 3D field, that will have already moved on before it gets here.
Keep well and stay safe.
Steve.
'AI is going to be our biggest game changer'
It most certainly is.
Looks up GPT-3.
I think we'll soon have an GAN that can take a CG image and "realify" it...
In fact, there's really nothing standing in the way of that being developed now.
We've only had Neural Network AI since 2012. Before that we were barking up the wrong tree with AI by using instruction lists instead of letting machines learn.