Robots will come to do human work. That was guaranteed. The general assumption is that they will take on manual labor, lifting heavy pallets in the warehouse, and sorting recyclables.
Now significant advances in production artificial intelligence which means robots are also coming to artists.
AI-generated images, created with simple text prompts, are winning art contests, make up, decorate Book coverAnd advertisement for “Piece of Nuts,” make human artists worry about their future.
Threats can feel very personal. An image generator called Stable Diffusion has been trained to recognize patterns, patterns, and relationships by analyzing billions of images collected from the public Internet, along with text describing the content. their content.
Among the images it trains is the work of Greg Rutkowski, a Polish artist who specializes in fantasy scenes featuring dragons and magical creatures.
Seeing Mr. Rutkowski’s work with his name enabled his stylistic tool to be effective enough that when Diffuse Stable was released to the public last year, his name become shorthand for users who want to create dreamy, fanciful images.
One artist found that whimsical AI selfies comes from the viral app that Lensa already has spooky signature on them, mimicking what the AI has learned from the data it has trained: portrait artists sign their work. Mr. Rutkowski said: “These databases were built without any consent, permission of the artists.
Since the generator came out, Mr. Rutkowski said he’s received far fewer requests from first-time authors who need covers for their fantasy novels. Meanwhile, Stability AI, the company behind Stable Diffusion, recently raised 101 million USD from investors and is currently valued at more than $1 billion.
Computer science professor Ben Zhao said: “Artists are afraid to post new artwork. Prof Zhao said: Putting art online is how many artists advertise their services but now they have “the fear of raising this monster that is becoming more and more like them”. “It shuts down their business model.”
That led Professor Zhao and a team of computer science researchers at the University of Chicago to design a tool called Ice cream on cake aims to prevent AI models from learning a particular artist’s style. To design the tool they intend to make Available for downloadThe researchers surveyed more than 1,100 artists and worked closely with Karla Ortiz, a San Francisco-based painter and illustrator.
For example, let’s say Ms. Ortiz wants to post new work online, but doesn’t want it made available to the AI to steal. She can upload a digital version of her work to Glaze and choose an art form different from her own, such as abstraction.
The tool will then make changes to Ms Ortiz’s artwork at the pixel level that Stable Diffusion will associate with, such as Jackson Pollock’s splashing blobs.
To the human eye, the Glazed image will still look like her work, but the computer-learning model will take on something very different. It is similar to a tool that the University of Chicago team previously created to protect photos from facial recognition system.
When Ortiz posts her Glazed work online, an image generator trained in those images won’t be able to mimic her work. Instead, a prompt with her name would lead to images in some sort of hybrid style between her work and that of Pollock’s.
“We are withdrawing our consent,” Ms Ortiz said. AI generation tools, many of which charge users to create images, “have data that doesn’t belong to them,” she said. “That data is my artwork, it’s my life. It’s like my identity.
The team at the University of Chicago admits that their tool does not guarantee protection and could lead to countermeasures by anyone committed to mimicking a particular artist. “We are pragmatists,” said Professor Zhao. “We recognize there could be a long delay before laws, regulations and policies catch up. This is to fill that void.”
Many legal experts compare the debate over the free use of artists’ work for AI to concerns about piracy in the early days of the Internet with services like Napster for allowing people to enjoy music without paying. Innovative AI companies have faced a similar set of court challenges.
Last month, Ms. Ortiz and two other artists filed a class action lawsuit in California against companies whose art creation services, including AI Stabilization, claim copyright and publicity violations.
“The allegations in this lawsuit represent a misunderstanding of how general AI technology works and the laws surrounding copyright,” the company said in a statement. Stabilizing AI is also sued by Getty Images for copying millions of photos without a license. A company spokesman said: “We are reviewing the documents and will respond accordingly.
Jeanne Fromer, a professor of intellectual property law at New York University, says companies can have strong fair use arguments. “How do human artists learn to create art?” Professor Fromer said. “They usually copy things and they are consuming a lot of existing artwork and learning patterns and pieces of style and then creating new artwork. And so, at a certain level of abstraction, you can say that machines are learning to make art in the same way.”
At the same time, Professor Fromer said, the purpose of copyright law is to protect and encourage human creativity. “If we care about protecting a profession,” she said, “or we think that creating art is important to our people as a society, we may want to protect artists.”
A nonprofit called the Conceptual Arts Association recently raised over $200,000 through GoFundMe to hire a lobbying firm to try to convince Congress to protect the intellectual property of artists. “We’re up against the tech giants with unlimited budgets, but we’re confident that Congress will recognize that protecting intellectual property is the right side of the debate.” the founders of the association, Nicole Hendrix and Rachel Meinerding, said.
Raymond Ku, a professor of copyright law at Case Western Reserve University, predicts that art creators, rather than simply collecting artwork from the internet, will eventually develop some sort of “legal system”. Private contracts guarantee a certain level of compensation to the author.
In other words, artists can be paid a nominal amount when their art is used to train AI and inspire new images, similar to how musicians get paid by streaming music companies.
Andy Baio, a writer and technologist check training data used by Stable Diffusion, says the services can mimic an artist’s style because they see the artist’s name alongside their work multiple times. “You can remove names from a dataset,” says Baio, to prevent the AI from explicitly learning an artist’s style.
One service seems to have done something along these lines. When Stabilize AI releases a new version of Stable Diffusion in November, it featured a notable change: the “Greg Rutkowski” prompt no longer works to get his style image, an evolution note by the company’s chief executive officer Emad Mostaque.
Steady Diffusion fans were disappointed. “What did you do to greg,” wrote one on the official Discord forum, which Mr. Mostaque frequents. He reassured the forum’s users that they could customize the model. “Greg training won’t be too hard,” another replied.
Mr. Rutkowski said he plans to start his Glazing work.
