The AI ​​shock – These jobs are already becoming obsolete today

"Imagine Mona Lisa, but with the face of Lionel Messie," wishes GabboDSQ. Others immediately pick up on the idea, writing "/imagine a chicken as Mona Lisa" or "A zombie as Mona Lisa". The Discord chat program is usually used primarily by computer gamers to coordinate their rounds together. But in the Midjourney project's Discord chat, users meet to play with art.

With the command "/imagine" - in German "Imagine" - the participants of an artificial intelligence (AI) called "Midjourney" give an idea via chat. This sets to work and draws the footballer Lionel Messi as Mona Lisa, in the style of the Renaissance painter and scholar Leonardo da Vinci. The AI ​​puts the result back into the chat as a digital image, the users can develop it further with simple commands, request variations, change details. The AI ​​presents a flood of options every second.

Midjourney is one of several projects by AI researchers that will transform the world of fine art, graphic design and illustration. Because suddenly users can let the computer draw. Quite simply, with a voice command in a chat program, without having to pick up a brush or touch a graphics tablet.

Midjourney and the competing projects called "Dall-E" and "Stable Diffusion" never tire, they never run out of ideas, they can imitate almost any style and know almost every cultural cross-reference. With serious consequences for entire professional groups, but also new opportunities for creative people.

"When I first came into contact with images that had been created with an AI, I found it rather threatening," says illustrator Christian Schlierkamp, ​​board member of the professional association IO, the illustrators' organization. "I'm thinking with the pen in my hand, and now a computer comes along and kind of takes the pen away from me, so I felt personally affected." For freelance illustrators, AI art came as a shock. The programs were publicly available for the first time in the summer – since then the industry has been in turmoil.

"On the one hand, illustrators from certain areas, such as the gaming industry, can easily work three to six times faster than before with the help of AI," says Schlierkamp. "On the other hand, according to experts, there is a potential loss of 60 percent of all jobs in other sectors by the end of the decade." Especially for illustrators in the field of advertising, mood pictures for book covers or online graphics, there was suddenly enormous competitive pressure.

Designers who design characters or worlds for computer games or film storyboards, on the other hand, can develop ideas much more easily than before thanks to the AI ​​or have entire worlds drawn in hundreds of variations - based on a few hand-drawn image specifications that they submit to the AI ​​for training.

The artificial intelligence comes from Elon Musk's start-up OpenAi (Dall-E), from various AI start-ups (Midjourney) and from the work of a lively open source software community based on an algorithm from the Ludwig Maximilian University Munich (Stable Diffusion). They were trained by their makers with billions of images from freely accessible databases - but also with photos from users of various social networks.

But how far do the possibilities of AI go? Does she just copy what her creators found somewhere on the internet – or does she create new ideas herself? In an experiment by WELT AM SONNTAG, Midjourney drew dystopian, futuristic images of dark islands in front of a city skyline from the specifications of a marketing text from the manufacturer Apple for the new iPhone, the smartphone hovering over them as a constellation of stars in the night sky.

A graphic designer would have paid dearly for the idea – Midjourney spat it out in half a dozen variants for free. The specification "Mermaid swims with orcas" produces a result that could also hang in a modern art exhibition, it looks Japanese-inspired - the vast majority of all museum visitors would probably not be able to tell the difference from a hand-painted picture.

"AI can synthesize photorealistic or graphic results and thereby creates something new, but that is not magical or outrageous, but a brilliant engineering achievement," explains Reinhard Karger from the German Research Center for Artificial Intelligence (DFKI) in Saarbrücken. "AI is creative in a technical sense, actually generative, but cannot create art, and therefore the mechanical 'creation' has no height." Art in the true sense does not arise when Midjourney draws, says Karger. "Because art is an expression of intentions anchored in everyday life."

The life experience, the will to want to express something, a computer just doesn't have. But what if people use the machine as a tool of expression, like a brush, telling them what to paint? Then computer images can also be art. Already, AI artists are discussing the best ways to tell the machine exactly what to paint - in the style of a particular painter, photorealistic or not, with strict format specifications or mixed as creatively as possible. They keep the best of their chains of command secret for fear of imitators.

All of this is little consolation for the artists whose style the programs copy. "Some in-demand modern artists, like fantasy illustrator Greg Rutkowski, are already at the point where their originals are getting lost online in the flood of AI images that imitate their style," says illustrator Schlierkamp. "So far, no one has been compensated for this."

The “haveibeentrained” page knows 5.8 billion images that are already being used for AI training – including the photo that shows the author of this text on the WELT.de website. “This is not legally permissible under European law. There is at least a lack of information for those affected. You would probably also need the consent of the person portrayed,” says Marc Störing, partner and data protection expert at the law firm Osborne Clarke. "Completely independent of this, there is still the issue of copyright, i.e. the use of the photograph and the copyright or ancillary copyright of the photographer."

For the first time, a robot answers questions in the British Parliament. Ai-da, an "ultra-realistic humanoid robot artist" developed at Oxford University in 2019, told members of a House of Lords committee how it works.

Source: WORLD | Lore Schulze-Velmede

What should an artist get if an AI imitates their style? What about a company whose trademark is used? During the test, Midjourney also accepts requests such as "Coca Cola on the Mars", drawing branded Coke cans without the company being able to give its consent.

"Here you should have a dialogue with the producers and the collecting societies," says DFKI expert Karger. He compares the process behind AI visual art to remixing music. "However, the AI ​​either makes the artists unrecognizable, and then they understandably feel negated," he explains. "Or the synthesized images resemble known styles, so that the creative work can be attributed, but then used without a corresponding copyright levy."

"Everything on shares" is the daily stock exchange shot from the WELT business editorial team. Every morning from 7 a.m. with our financial journalists. For stock market experts and beginners. Subscribe to the podcast on Spotify, Apple Podcast, Amazon Music and Deezer. Or directly via RSS feed.