The Dangerous Economist: Will technology cost artists their job?

Monday, January 09, 2023

Will technology cost artists their job?

Structural Unemployment-unemployment caused by a mismatch between the skills of job seekers and the requirements of available jobs.

One example of this is when you are replaced by a machine. Another example is when there is a fall in demand for your product, so you get laid off, like with typewriters since people now use computers. A third example is geographical, when the jobs are not in your region of the country.

See If you can type it, you can see it: Text2image AI turns words into pictures by Dwight Silverman of The Houston Chronicle. Excerpts:

"On Craiyon, visitors enter words to generate images that are sometimes realistic, sometimes whimsical and occasionally disturbing."

[this] "creation is known as text2image. While Craiyon’s results clearly aren't realistic, other models can output images so advanced that it’s hard to tell them from photographs or human-created art, generating both excitement and anxiety.

In the art community, some worry such technology will cost artists their jobs. But others see it as an opportunity to inform and enhance their work."

"There are myriad AI models with common interfaces: Users enter text and the model generates images based on those words. Craiyon offers nine images in about a minute.

Craiyon’s model is trained and its images produced, Dayma says, using data from “hundreds of millions” of pictures, captions and metadata scraped from the web."

"GAN, or generative adversarial networks, is an older AI approach trained to output a specific image based on other images. An example of a GAN model can be found at www.thispersondoesnotexist.com. The site, which launched in 2019, generates a realistic human face each time the web page is reloaded. No human matches the face – each is fake – but the site raised alarm bells when it appeared, making it clear AI has progressed to where almost no image can be trusted to be real.

Models known as diffusion and transformer work from a text description. The words are applied to the dataset of images learned by the AI, and one matching the text is eventually constructed. Craiyon uses the transformer model."

"In August, the Colorado State Fair awarded first prize in its art contest for a work that was created using AI. The artwork, depicting a portal opening inside an ornately detailed room, was submitted in a category for digital art. Jason M. Allen told The New York Times he was open about how the piece was created, and the judges stood by their decision. But the award incensed some artists and created a social media backlash. Some artists are concerned about AI’s ability to mimic styles. An artist’s style can't be copyrighted."

"And coming soon: Text2video. Facebook parent company Meta this month announced Make-a-Video, which generates a few seconds of video from a text prompt. Not to be outdone, Google announced Imagen Video. Neither is available to the public."

"You can see where AI-generated art is headed in the work of Mary Flanagan a Dartmouth College art professor, game designer, programmer, writer and author. Flanagan, who has a home base in Houston, has coded her own AI designed to create images, Grace:AI, and trained it on tens of thousands of pieces of art created by women.

Though some artists are upset about AI-generated art putting some of them out of work, Flanagan isn’t among them. She says AI art will free artists and designers “from the really boring work” so they can concentrate on their own creativity."