NEW DATA POISONING TOOL LETS ARTISTS FIGHT BACK AGAINST AI - IS THIS A GOOD IDEA?
― Alan Turing, Computing machinery and intelligence
NEW DATA POISONING TOOL LETS ARTISTS FIGHT BACK AGAINST AI
This was shared high and low by jubilant artists, but keep this in mind:
"Zhao admits there is a risk that
people might abuse the data poisoning technique for malicious uses."
What could it mean? Well, for one thing, it is not entirely impossible for a mischievous nerd that lives in his mothers basement to download and tinker with this technology, pick any artists name at random, and poison it.
The results would be, that instead of a cow jumping over the moon, you get a handbag floating in space. That would be a tame result, but what if it is something more sinister? Such as some fascist symbolism or whatever comes into this devilish mind.
Even as an artist itself using this tool, you cannot be sure of the outcome.
In many AI applications the prompts are open (the option of hiding prompts is available in some), so you, the artists name, is prominently displayed as a modifier of the prompt. Would you honestly want to be associated with the results? Results that are not in your style or mode of painting, results that could at a minimum be embarrassing when associated with your name? Because your name is on it! The one who generated the AI would not be that concerned about it. In fact I read one comment that said (s)he looked forward to such crazy results. But would you? The 'creator' of the AI piece is mostly a unknown, but you, the artist, whose name appears on the prompt, is possibly well known!
I spent a lot of time in the academic ivory tower, and I came away with the conviction that most of them live in their own little box of knowledge, oblivious to what goes on outside of this bubble. So they come up with some stuff like this that once released and out in the wild, it is like opening Pandora's Box. In other words, you cannot put the genie back in the bottle. They are like the Sorcerer's Apprentice, to drive the analogy further.
Artists are of the impression that AI is all about them. But in fact the application to art is just a small side track. There are uses in Science and Industry, Medicine, Research, Museums, Search Engines, sophisticated programs like Adobe Photoshop (to name just a few) that rely on the use of AI as a tool. What would happen to that if this poison spreads? It took years for AI to become the sophisticated tool it is today. A poison program that only existed for a few months could very well grow just as exponentially and be a real menace in the mainstream of AI application.
all images except the top one are my AI creations - the last one here links to 'unabashed' self-promotionBut yes, according to the haters, it is all about art and (the ego) of artists.
But I outlined above what it could mean to you, the Artist, in the first place.
Visit my website
▼
PRINTS OF MY ARTWORK AVAILABLE HERE:
Bonus Special: JOURNEY THROUGH THE WORMHOLE
should you not be on the blockchain, or have no upvote power, or this post is already older than 7 days, you can always just buy me a coffee.
ai
nightcafe
art
surrealism
datapoisoning
technology
machineintelligence
opinion
creativecoin
science
0
0
0.000
Are those tool not applied when the artist uploads a new piece of art? How can a trickster exploit this? By flooding the internet with poisoned copies of the art? But are the models really trained by grabbing any images from the internet? Don´t they have distinct sources?
Training models is a ongoing thing, that is why you have so many new modes coming out, while old modes are discarded. But since this supposedly came out last October, I have not seen any adverse effects with the various modes and prompts I use. The part that is effective in some way is the option for artists to opt out - in fact I seen this particularly on Deep Dream Text2Dream: the pick list of artist modifiers has shrunk a bit.
But I think that this 'poison' is in infancy, and could possibly be altered and expanded in a malicious way. Of course if you download the app you swear on a bible you won't screw around with it, but would such agreement bother a hacker?
Just because the developer of the program sees it as unlikely to happen, doesn't mean it couldn't, all it tells me that he cannot see beyond the edge of his teacup.
I don't think Nobel had initially any bad thoughts about his invention, he just wanted it for defense. Changed his mind rather quick.
regarding a artist using the poison app on new uploads ... I also said this:
In many AI applications the prompts are open (the option of hiding prompts is available in some), so you, the artists name, is prominently displayed as a modifier of the prompt. Would you honestly want to be associated with the results? Results that are not in your style or mode of painting, results that could at a minimum be embarrassing when associated with your name? Because your name is on it! The one who generated the AI would not be that concerned about it. In fact I read one comment that said (s)he looked forward to such crazy results. But would you? The 'creator' of the AI piece is mostly a unknown, but you, the artist, whose name appears on the prompt, is possibly well known!
About the same text on Facebook, after a long discussion,I said this:
Quote: "instead of bitching about AI, I educated myself about how it works. I am far from being an expert like those who could get SD to make exactly what they want, but I learned enough to be able to train my own modes. So far, done two. One with my drawings and one for selfies. They are private, nobody else can use them. I plan on adding a couple more, one with my paintings and another one with a specific painting technique I had been using for decades. SD does not have any training on the later, since when I input a prompt about it, it comes up blank. Even with a seed image. I hope I succeed, I just need to gather enough images together. Like, for my drawings I used over 60 to train the mode. Once done, that too will be private.
Furthermore, a few times I had been using Photoshop together with AI generated images, not just for corrections, but for entire new digital creations, combining images, layering, distortions etc. Endless possibilities.
Next I could be using such images as input seeds. In most cases I use a prompt with my name and about myself, sort of a summary of who I am. The results are totally myself."
Hello thermoplastic!
It's nice to let you know that your article will take 14th place.
Your post is among 15 Best articles voted 7 days ago by the @hive-lu | King Lucoin Curator by nuthman
You receive 🎖 0.6 unique LUBEST tokens as a reward. You can support Lu world and your curator, then he and you will receive 10x more of the winning token. There is a buyout offer waiting for him on the stock exchange. All you need to do is reblog Daily Report 187 with your winnings.
Buy Lu on the Hive-Engine exchange | World of Lu created by szejq
STOPor to resume write a wordSTARTThe post is already past 'Best before date'. But here is a interesting addendum to this:
On Facebook there are so many 'Experts' that rant about AI, in posts and with comments on posts, so I posted this item, with this text added:
While so many artists get their panties in a knot over AI, here is a high level case of clear copyright infringement that took almost 8 years to come to a judgement. Involved are some of the top galleries in New York, such as Gagosian.
And guess what? Dead silence! I even tagged some of those people, but nobody said a word.
Österreicher die sich interessieren - Austrians that are interested
hier ist ein Link über KI Bildung - here is a link about AI schooling
https://www.bmbwf.gv.at/Themen/schule/fpp/infomail/2023_11/ki.html