Image generators: How the artist community can now defend themselves against AI art thanks to "Glaze"

The research group at the University of Chicago has now developed an effective technique that enables artists to protect their works of art from AI image generators who use these artificial intelligencies to create images in an existing style.

As reported by the New York Times software is called "Glaze" , which places a kind of "veil" over the pictures that is not visible, but prevents AI systems such as Stable diffusion from reproducing the style.

This approach is reminiscent of methods used to prevent facial recognition.

How AI image generators shake the art world

It is not an exaggeration to say that the introduction of text-to-image generator models caused the art industries to shake and triggered a real trigger among artists. By sending simple text demands such as "A picture of a dog on the moon" to processing models such as Stable Diffusion or Midjourney , every user can suddenly create incredibly detailed, high -resolution graphics and works of art.

Before that, many working hours from professional artists were necessary. AI art even gained prices on established art congresses, served as cover pictures for magazines, and is used to illustrate video games

More powerful models continue to come onto the market, which are catalyzed by VC financing and technical research breakthroughs. Just a few months after their provision, these models quickly arrived with a wide user base and grow rapidly in user numbers and platforms. In September 2022, Midjourney reported over 2.7 million users and 275,000 AI works of work that are created every day.

Beyond open legal questions of copyright, intellectual property and artistic consent to use, it is clear that these AI models have significant negative effects on independent artists. For the estimated hundreds of thousands of independent artists
all over the world, most work on order and attract customers by advertising patterns from their works of art online.

Initially, professional artists undergo years of training and develop their individual artistic style. The AI ​​model is also based on imitation of existing styles and benefits from training with them without compensating for the responsible artists. This effectively ends their ability to make a living.

Since the imitation of synthetic art continues to increase in popular artists, they displace original works of art in the search results, which further impaired the ability of the artist to present and present work to potential customers.

Finally, they demoralize art students who develop into future artists. Art students see their future careers replaced by AI models.

Fear of marketing your own art

Artists are afraid of publishing new work on online platforms, Ben Zhao, a computer scientist who is involved in the project, told the US newspaper. They fear that they "feed the monster that is getting bigger and bigger" .

This would destroy your business model. For this reason, you have developed a tool with which you can provide works of art to download that are slightly modified to make it impossible for AI technology to recognize a style. A test with 1000 artists has shown that Glaze is extremely effective .

It was quickly obvious that AI image generators such as Stable could be used diffusion or Midjourney to imitate the style of individuals after they have become successful last year. The case of the Polish digital artist Greg Rutkowski , whose style was often copied, made this clear.

He stated to the New York Times that he "significantly fewer inquiries" regarding illustrations of his works. This criticism had also arrived in those who have developed Stable diffusion and therefore the AI ​​no longer generated art in the Rutkowski style.

The conflict escalates

The team from Chicago explained that artists can upload their work to Glaze and choose a different style, which means that they are changed minimally and can no longer be used as training material for AI systems. This enables them to be put online without having to be afraid that a AI learns the style.

However, they admit that this is not future -proof and countermeasures can be found at some point. In the meantime, however, new works can be protected. A Windows and Mac version are planned, the broad distribution is hoped for. Rutkowski wants to use Glaze .

After last year's turmoil around the AI ​​image generators , the software supported the creators to effectively defend themselves against the non-generated use of their material. In the past few weeks, lawsuits were filed against some AI companies .

Raymond Ku , a copyright expert, told the New York Times that he expected that a kind of compensation system will be developed in the near future, from which the artists may receive minor financial compensation.

The conflict should continue to increase up to this point.