words matter to me crazywabbit :D
a great script will support poor art
but brilliant art will never salvage a dire script
so says a cartoonist....
Printable View
words matter to me crazywabbit :D
a great script will support poor art
but brilliant art will never salvage a dire script
so says a cartoonist....
You are so wise handrawn. ;))
l'd leave the ironic sarcasm to those who are good at it, otherwise some might think you are trolling :p :D
Alright girls. Enough bickering.
What is scary to me contrary to what Steve is saying is the ability of AI to create a scene on it's own with minimal input complete with details that we not in the request.
To achieve this degree of detail through description would require several pages of description. The actual description for the scene (not my image) was minimal. The AI bot drew upon a huge database of visual information including different surfaces, light sources, time of day, and a lot more. Pretty much the same process a skilled illustrator or painter would.
gary, the wonderful image produced is of limitrd use to a designer unless it was what they actually wanted, where is the artistic control if it starts adding things you didn't ask for; you get clipart like I said - wonderful stuff, but if it did it for you, it might do exact same for someone else; clipart
this will be a tool for liberating artists so they can work at another level, like a computer did for tech drawing..
BTW I am only a girl when I choose to be....:p
...
Aren't we all.Quote:
BTW I am only a girl when I choose to be....:p
Having worked in advertising for 15 years and then having my own graphic design company for another 12, I know how much damage this is going to do to the art and photography business.
A client is going to ask, why are we paying (and way back when I was working) So and So $15,000 for an illustration/photograph when we can get an image for the price of an account (chump change) from one of the AI graphics generators.
And for that matter, why are we paying So & So What all that money and benefits to create advertising when Chat GPT can do the same for peanuts. And why are we paying all that money to the research director, the account executive, the entire creative department.
I know how the business works. And this is not the small shops. I worked for Young & Rubicam and other major international ad agencies. Turn them upside down and they all look (and think) alike.
I'm with handrawn. I have yet to have any of the generators produce something from my prompts that is close to what I'm envisaging. Until we get prompts, ChatGPT style, then AI art is very limited if you're looking for something, even loosely, specific.
With ChatGPT I set out to make a slider that had individual links and scrolled right to left. It produced code in seconds. But the slider reached the end and then flicked back to the beginning. I carried on with the same question/prompt by asking to have the slider reverse when it reaches the end. Again, mere seconds for the code. Lastly, the slider had a double time pause at the first and last slides. A continuation of the prompt, and I now had what I was envisioning.
To be truly job stealing, AI art generators would have to work in a similar manner. Imagine a client/manager asking for a stylised logo, side view of a woman's head with flowing hair. You slap the exact prompt into a generator and email the result to the client. "Make the eyes slightly wider, make the hair longer and more flowy, and the lips look like she's pouting." ChatGPT style you could ask for the changes. As it is, you're out of luck.
gary - yes and when we get control that control will be available to clients as well and they may choose to do their own designs
just as currently clients can choose to use tools (I"m looking at you xara) to do their own corporate branding and some their own animations
this is empowering, but design studios will have to adapt or, I guess, go the way of hand drawn cell animation
chris - the controls may be a way off, really it needs a ui interface, i'd suggest calling it blender but someone has already thought of that
and then for nontech clients something a bit simpler
it may take a while...
Indeed Gary but here's the failing. The request probably required a brass & steel robot, a clock, dinner table, brass table lamp chandelier. And the results at first glance look impressive but as you look closer you start to notice the oddities. There are three clocks, two on the wall next to each other which would be very strange in the real world. AI tends to draw clocks at 10 past 2 because that is how the majority of clocks in it's database are set. Ask it to draw one showing a specific time and it has no idea. The table lamp on further inspection looks like something Esher would create. The four and a half oddly scattered equally filled wine glasses look very wobbly etc.Quote:
...create a scene on it's own with minimal input complete with details that were not in the request
It becomes very complicated if you want to create a cartoon comic strip say of a consistant character where you need to use seeds but these often wander off the instructions often replacing a white baseball cap with a lue one for no apparent reason. It then becomes very time confusing.
As for taking over jobs, computers have been doing this at ever increasing speed over the years mainly in the manual sector. Look at the print industry alone. It is unrecognisable from that of 50 years ago. It's now moving into more complex work.
nicely put egg
We are still in the early stages of AI created art.
But Egg is correct about the clocks. Here are two attempts to get all clocks to display 9:37
And as well as often having the wrong number of hands and big issues displaying the roman numeral IV for 4 o'clock for some reason whilst managing IX for 9 without much problem.
However it blows my 10 to 2 theory to dust.
I had to add the 10s & 6d label!
New Data Poisoning Tool Lets Artists Fight Back Against Generative ai
Article on Slashdot that was very interesting. Thought it might fit in with this conversation.
https://slashdot.org/story/23/10/24/...-generative-ai
sunland
Love it. Thanks for posting from your fellow Nuevo Mexicano, Bookmarked. — Gary
That's refreshing news Sunland.
I wonder how soon this will be available. It sounds great though of course there has been so much material scraped already.
Glaze, developed by the same team that is working on Nightshade have a product called Glaze, in beta that does much the same thing. 500MB file size. https://glaze.cs.uchicago.edu/what-is-glaze.html
If you read the article and the considered Comments then I wouldn't hold your breath. The processes have not yet been peer reviewed; others have already worked out counter strategies; the billions of images so far scraped will not be affected; you may just be adding noise to your image; lawyers have already argued there is no copyright infringement, citing fair use.
If you were to go down this rabbit hole and present using a Xara product, you would have to check every image section is covered by it or has a watermark. Quite difficult, especially if using HighSlide.js.
Acorn
Is it fair use to scrape every one of Stephen King's novels? Or everyone of Margarete Atwood's novels?
Fair Use is a very weak argument and generally refers to non-commercial use. I see an image of yours on the web, I download it for a class project. But if I download that image of yours and repackage it as my image — even with a few modifications (this happens all the time with our images being offered as NFTs) then that is a copyright violation.
But yes, these apps amount to locking the barn door after the horses have been stolen. It would have to be going forward. If there is any going forward at this point.
I'm not taking sides here, just adding food for thought.
The scraping of existing work, be it art, science, coding or literature, isn't the hill to make your stand on. Stephen King learnt to read and write by (essentially) scraping knowledge from people before him. I don't know in Mr King's case, but a lot of creators will say they were influenced by other people. The problem that will need addressing in the near future is precisely replicating existing works.
Two real world examples.
There are AI voice changers using famous singers. My son sent me a track he was working on, saying that Taylor Swift had emailed him begging to do the vocals. He even sent a sample. It was, indeed, Taylor Swift...AI Taylor Swift.
Mr Trump has been popping up all over social media in all sorts of places, poses and scenarios that the real Mr Trump would not be seen in.
Both of those examples are here, today. Both will fool the vast majority of people. How long will it be before lobbyists, lawyers, politicians, corporate HR etc. are caught proving that Mr Trump actually was drunk as a skunk on the number 77 bus singing like Taylor Swift. We have video proof. Here's a picture. We have audio.
Anyone can look at a painting and emulate the style, adding their own 'flair'. Anyone can read Mr King's novels and write about a possessed car. As long as they don't claim to be that actual someone, then that's acceptable. Why is AI any different?
I genuinely think that vandalising an LLM is rather petty and completely pointless. Anyone remember dual cassettes, VCRs, DRM, DVD burners, 3D printers. All of them were going to be the ruin of something.
now it my turn to be with Chris
It's a game changer - the old 'good artists copy, great artists steal' with a new twist, and I don't think you can hold back the tide
a question of degree... and we may have to accept that, up to a point, we have left [strict] copyright behind; in the same way that you can legitimately manipulate a person's image as long as you do not defame them, you can now do the same to their creations as long as you don't replicateQuote:
Anyone can look at a painting and emulate the style, adding their own 'flair'. Anyone can read Mr King's novels and write about a possessed car. As long as they don't claim to be that actual someone, then that's acceptable. Why is AI any different? [cf:] The problem that will need addressing in the near future is precisely replicating existing works
that is what is meant here by 'fair usage' - sampling has been part of commercial music for quite some time; this is sampling on a grand scale...
>>> Making Chat (ro)Bots - YouTube <<<
8 minutes long. Shows how changing the FDM (Fine Detail Model) is just as important as the prompt. Same robot, same simple prompt, different personalities (FDM).
I've found this to be true when creating AI art as well.