By Cameron Butler
Editor-in-Chief

A remarkable new development in the technology sphere stands to disrupt billion-dollar industries and countless career fields. Students at Harold Washington College may not be unaffected by a new “singularity” as they approach the workforce. This is thanks to the emerging field of AI-generated content, a potential beginning-step into the Imagination Age.

“A cookbook photo of a salad pizza” generated with DALL-E 2

An oil painting of a cheeseburger. A daguerreotype photo of a colonial woman on Mars. A Nike Air shoe in the style of a clown. A 3D emoji of a nerdy alien. A photo of a palm tree in Antarctica. A cookbook photo of a salad pizza. A medieval painting of a cat wearing sunglasses. A child’s drawing of a couple standing under a rainbow. An x-ray scan of a guitar. A painting of a woman dancing with a snail.

“A pencil sketch of the Mona Lisa” generated with DALL-E 2

At first glance, these prompts may look like the search history of a manic dilletante, but they represent a disruption to reality as we know it. That’s because none of the images originating from these prompts are real; there was no camera that took the photo, no brush that painted the image, no computer graphics software that rendered the shape. The pixels were all dreamt up by artificial intelligence text-to-image systems like DALL-E 2, Stable Diffusion, and Midjourney. Imagine a world where you will no longer truly know if what you’re seeing on your screen is real. That world is here.

Text-to-image models work by taking a text input and turning it into a novel image through training on billions of existing images, without re-using any of the pixels in the dataset images.

“An x-ray scan of a guitar” generated with DALL-E 2

Text-to-image can be thought of as an advanced form of text-to-speech. Instead of simply converting text into spoken words, text-to-image models create a visual representation of the text. This most commonly involves using a computer algorithm to analyze a text prompt, which then generates an image that corresponds to the words and concepts within the prompt. This process can be compared to how a child might learn to associate different words with different images, like learning that the word "apple" is associated with a picture of an apple.

77% of students and staff polled at HWC in September had never heard of text-to-image programs, otherwise known as “AI art”.

Of the remaining 23% of respondents who were familiar with AI art, 42% of those knew of DALL-E 2, 28% of Stable Diffusion, and 14% of Midjourney, the three most popular text-to-image systems to date.

“I think AI art is really cool in theory. I think it's interesting to see how computers and AI understand and interpret years upon years of art from every corner of the earth,” said Crow Cameros-Lopez, a 20-year-old Architecture student at HWC.

“I think it’s a pretty incredible invention. A scramble of words that the AI can create into art, it’s pretty neat,” said 25-year-old student Alejandro Soto. “I think it should be incorporated into schools to help students get into their minds and help figure out projects, or even hobbies for themselves.”

“A child’s drawing of a couple standing under a rainbow” generated with DALL-E 2

“AI art” made headlines earlier this summer when a piece generated with Midjourney was selected as a winner in the Colorado State Fair in the digital art category. The first-place prize was $300, which was on top of the $750 the work later sold for, marking the first time an image generated by artificial intelligence won an art competition.

Whether or not AI itself can create art is currently up for debate among those knowledgeable about the emergent technology.

HWC Associate Professor and Digital Multimedia Design Coordinator Galina Shevchenko described artificial intelligence models as tools, rather than artists.

“I think the AI program allows for [artistic] discovery. AI is a tool, I think. It can be used as a tool. And I know some artists are using that as elements of what they are doing, but essentially, it's not like AI is a creator; the artist is a controller. The artist is somebody that controls that AI algorithm,” said Professor Shevchenko.

“I would not describe it as real art. As cool as it may be, in theory, it's missing what I think separates art from being real or not,” said Cameros-Lopez. “AI art doesn't contain the emotional impact, the culture, the meaning, or the purpose, in my opinion, as real art does.”

“A Nike Air shoe in the style of a clown, product photo” generated with DALL-E 2

The advent of these AI systems may mark the beginning of a new Age of Imagination, a proposed successor to the Information Age, where ideas and concepts begin to hold greater economic and cultural value than data and information itself.

“These tools, they're not just substituting for a human, they’re becoming human prosthetics. They are becoming an extension of our intelligence because we can take them and use our intelligence to tell them to do stuff that we want to happen,” said Professor Shevchenko.

“A 3D emoji of a nerdy alien” generated with DALL-E 2

“All of this technology only benefits these areas, rather than substitutes. It’s like us learning all these languages and trying to write poetry with those languages,” said Professor Shevchenko. “It's not like it's a button-pressing AI thing. It's something very conceptualized, something so poetic, that AI is just the tool to implement those concepts.”

Getty Images announced in late September that they would “cease to accept all submissions created using AI generative models,” explaining that the copyright status of these images is currently not well-defined.

Earlier this year, the United States Copyright Office denied the registration of a copyright claim to a work generated by artificial intelligence, stating that the work “lacked the required human authorship necessary to sustain a claim in copyright” and that “copyright law only protects ‘the fruits of intellectual labor’ that ‘are founded in the creative powers of the [human] mind’”.

“A watercolor painting of a woman with leaves for hair” generated with DALL-E 2

The Supreme Court has yet to rule on the specifics of whether or not an artificial intelligence can hold copyrights and patents.

“AI is only limited by what we've done. It's ‘stuck in the past’ and can't create new concepts and ideas like people can, and I think being brave enough to create the unthinkable is the magic behind art,” said Cameros-Lopez.

Students and staff at HWC were mixed on the potential downsides to this disruptive innovation.

“I think I'm really fascinated with all this new technology. I'm really fascinated how much it's at our fingertips. I'm a big fan of all this new stuff,” said Professor Shevchenko.

“AI art definitely brings me a bit of grief in terms of my future career. Knowing that one day it could become ‘big’ and what people think are the new NFTs. That thought does bring a little bit of worry, however I don't think it would impact my career too much, as I doubt it would get to a point where it would replace the craftsmanship of an actual person,” said Cameros-Lopez.

“A photo of a raging wildfire at the Pyramid of Giza” generated with DALL-E 2

Despite these concerns, the student process of learning will continue alongside the advent of this new form of media, according to Professor Shevchenko.

“In terms of the higher education, students that we teach still need to learn those digital tools. You need to know how to make those things work from scratch and then you understand how this AI is made,” said Professor Shevchenko. “Also, a lot of these AI programs are trained on real artists, so they're not happening from scratch. All those images that you see, it's because there were lots of digital files that were created by real artists that have been brought in into this AI program to generate those images. So, AI actually feeds on the real art.”

DALL-E 2, the text-to-image system from OpenAI, recently received $1 billion in funding from Microsoft for its development of AI tools.

Stable Diffusion, another text-to-image model, was in talks in September to raise up to $1 billion in valuation from potential investors, according to a report from Forbes.

“An arm tattoo of a clown” generated with DALL-E 2

“In AI, a lot of times you lose control because the AI takes over, right? So, you kind of have to set those parameters of how far you would allow yourself to lose control or how far you will let the machines take you. Or maybe you want to see what happens when it takes you somewhere further,” said Professor Shevchenko.

As of September 28, DALL-E 2 is available to the public, according to an announcement from OpenAI.

In a similar vein, GPT-3, another AI system from OpenAI, can generate text from user-inputted text prompts. A marketing slogan can be generated for a new flavor of soda. A moving poem can be generated from a list of emotions. An essay can be generated in seconds, if a student inputs the essay topic.

With these generative technologies now accessible to anyone who wishes to use them, educators may never know if what they’re grading was generated by an artificial intelligence, or by a student. Current plagiarism software has no way of detecting text generated by AI.

Seeing may no longer be believing.