AI—A Blessing or a Curse?

Fredrik Malmedal Salhus 

Teknologisk bilde

AI generated by: Midjourney. Prompt: ‘portrait of cyborg human in 2140 organic intricate detailed octane render photorealistic --v 4’.

I want to start off by saying that I am very ambivalent about the topic of AI generated art. Is it a blessing or is it a curse? I can’t seem to wrap my head around the question.

I both love and despise AI generated art. I don’t know whether I should laugh or cry when I get image results from an AI. If I manage to adjust my prompts to a description that yields high quality results, I get dazzled and amazed, but completely terrified at the same time. What are, and what will be, the consequences of the fact that we can generate content this way?

When I get disappointing image results from the prompt page, I feel misunderstood. I get this urge to fine tune my choice of words; to be more specific. It is as if the site is whispering to me: ‘Look at all these incredible results people are getting. You can also generate amazing images, you just have to try again—and again. ‘Come on now … spend your credits, explore more’.

The past six months I have been learning the prompt language, and I see it has a lot of nuance to it, both technically and descriptive. Nit-picking at words and implementing cryptic code to get desirable results from image generators like DALL·E, Midjourney, and Stable Diffusion seems to call for continuous or even endless practice.

I am both worried and enthusiastic for artists, designers, and the future of the visualizing industry. The momentum AI technology has now seems unstoppable. I am under the impression that the algorithms and the image results are improving as time passes. New iterations and updates of AI generators are constantly being released. Input from millions of users over the past six months has been rocket fuel for these AI ventures, launching it to the moon and probably further into space sooner rather than later. Optimised, tweaked, and improved by software developers and scientists, and the utilization of all of our valuable input data. Computationally learning to perform better. In a Youtube clip from the channel ‘Help Me Understand Clips’the Artificial Intelligence expert Dr. Thomas R. Jeffrey says;



Teknologisk bilde

‘The goal of machine learning is to learn from multiple iterations of something’

Let´s wind back the clock a bit. Last summer the visual communication industry was hit with a paradigm shift in technology. The release of online artificial intelligence image generators that create unique images based on text input called prompting. This means everyone can type nuanced image descriptive words into a text box and let Artificial Intelligence create unique digital artworks in just seconds.

How did that happen? It is a result of exponential development. Technology that could only generate a tiny unrecognizable 32x32 pixel image of a green blob seven years ago, in a lab, based on the text prompt ‘green apple’, has developed rapidly and can now generate unique images in seemingly all thinkable visual styles defined by the user.

The technology is manifold and is generating waves of innovation in multiple areas. Soon we will be able to generate 3D-files from single images. Free and open tools are already available to do the opposite as well. On the Replicate (2022) website we can generate descriptive prompt words based on images we upload.

The increasingly popular and well known open source 3D program Blender has new plugins that render artificially intelligently made textures on digital models automagically. And voluntary members of the community around Blender are currently developing more AI features using the open-source code made available. (Hawat, 2022)

In his book Profiles of the Future: An Inquiry into the Limits of the Possible, published in 1962, science fiction writer Arthur C. Clarke wrote ‘Any sufficiently advanced technology is indistinguishable from magic’. (Efraín Foglia et al, 2018)

Teknologisk bilde

AI generated by: Midjourney. Prompt: ‘portrait of Arthur C. Clarke with futuristic look in the style of cyberpunk octane render’..

The AI platforms surely felt this way to me when I started using them. The experience has a whiff of sheer magic to it! It is fun, exciting, and mesmerizing.

AI technology is, however, forcing change in all industries right now: It is coming for bus drivers, bankers, lawyers, editors, developers, publishers, journalists, photographers, shop owners, athletes, chess players, and now visual artists and designers of whom many thought they would not be impacted. (Rowan, 2019)

Already in September of 2021. Nearly a year before it all took off, the scientist and writer Eric Hole wrote;

‘In my office hangs a painting called “Rule of Consciousness.” An ironic title, as it actually signifies the end of the rule of consciousness. For it was designed by an AI. I keep it not because I like it. In fact, I hate it. Or rather, I’m afraid of it. It’s a sobering reminder of what’s coming, which is that human art is close to total control by corporations, and no one seems to care. Hoel. E (2021)

Humans are adapting to the ever more increasingly, optimised and complex landscape of computers outsmarting us in speed, accuracy. AI powered robotic systems are now running warehouses (Del Ray, 2022) and winning over all humans in games like Go, Chess and Poker (Wali, 2022). They consist of continuously fed optimized algorithms and code and it improves the technology in time in an exponential way. Technology advancement has been a part of industrial progress since the dawn of the industrial revolution. And the utilization of it has not been without push back. There were people in the early 1900s that didn’t want automobiles on the roads.

‘... the anti-car movement was a real thing that lasted until 1930, when the US reached more than twenty cars for every hundred people, or about one per family.’ (Scott, 2016)

They thought it would suffice to have horses and buggies, and they were probably right, before critical legislation and sufficient infrastructure was in place.

The way I personally observe the situation is that we are facing a faster rate of change than ever before. Due to digitalization, we are witnessing monumental changes occurring over much shorter timespans, and we the people have a hard time adjusting to the situation.

The infrastructure and systems that are being put in place now do have dire side effects as they increasingly pave way for a more authoritarian society.

The authoritarian part I refer to is the fact that our online presence is increasingly tracked and monitored. An extreme example is that social score systems implemented in certain cities in China are pointing at a bleak dystopian future of an oppressed civilisation.

It is, in a sense, perpetuating a trend which leads to a ruling elite over an otherwise technologically distanced population.

The nerds have become the new emperors.

Teknologisk bilde

AI generated by: Midjourney. Prompt: ‘nasa space astronaut riding on a rocket launching away from earth detailed photorealistic --q 2 --v 4’. Edited by: Fredrik M. Salhus

Do we need all this?

I have to ask:. Do we really need all these tools? I can, as an example, refer to Adobe Sensei (an new piece of software from Adobe) which artificially analyses and discerns all objects in images and gives you the opportunity to delete them, move them in 3D space.

I would argue that just because something is programmatically possible that doesn't make it a needed feature. It is impressive tech, but is it necessary? Chief Product Officer at Adobe, Scott Belsky, said in an interview recently ‘We want to benefit creatives, not replace them”’. (Piper, 2022) Mmmmhhhmmmmm. Right … But I guess what he wants is to sell automated AI generated stock photos, and not need to pay content creators their fair share anymore.

In my observation a seemingly increasing number of Youtubers are sharing the gospel in the AI echo chambers. They are telling their audience that they don’t have to be ‘professional artists’ to create amazing art and photos. ‘As long as you have good ideas and you can describe what you want, … Bingo! … here you go sir. Here is your piece of art.’ (PiXimperfect, 2022

Teknologisk bilde

Ethical issues

To me it just seems awfully wrong to reference (copy the style of) specific artists to create ‘your own artificial piece of art’. Especially when the licence agreement between the users and the platforms are diffuse about the ownership of the art you generate.

I understand the temptations of working ultra-fast, the high inspirational level, how efficient it can seem to generate references and documentation for research, but it makes me think more deeply about creative ownership and authorship. What is ownership? Physical possession is one thing. If you have something locked up, then it is out of reach. Digital works, however, are so easy to copy and transfer.

Being conscious of Bitcoin and one of its novel value propositions (impossible to copy) we now have ways of copyright protecting digital artworks. The invention of Bitcoin has paved way for digital scarcity. Before Bitcoin there was no way to limit people from copying digital tokens. One of the many aspects of the Bitcoin invention is the solving of the digital double spend problem. Sadly, sharing open-source code has led to the escalation of the NFT frenzy and a scammers paradise exposing a boom-and-bust tulip mania phase. This is currently under supervision from SEC as many tokens classify as securities and are traded on unregulated exchanges which in turn easily leads to a lot of wash trading, money laundering and other shady activity. (Chainanalysis, 2020)

I guess that, to some degree, all still images are copyable, because you can always take a picture or a screenshot of it to then recreate or restore it to higher resolution and present it as your own work.

¨
Teknologisk bilde

AI generated by: Midjourney. Prompt: ‘beautiful devil woman depicted in hell photography lighting + 50mm, 80mm, 100mm + lightroom gallery + behance photography --ar 2:3 --q 2 --v 4’.

Data scraping

Scraping is a term used for stealing artworks and images en masse from the entire world wide web in order to train and improve a for profit owned image learning platforms. They also improve algorithms in order for subscribers to be able to emulate any living or dead artist out there. I find this devious, fraudulent and see it as a stab in the back of every struggling artist out there. The community should rise up with pitch forks and claim payment from all these engines every time their name is being referenced and used in a prompt.

It does worry me that Midjourneys generated images look like a combination between things we have seen on Pinterest, Artstation, Behance, and in Dream Works, Disney and Pixar movies. I really do wonder; has the usage of all the millions of reference photos been addressed and clarified legally?

The latest scraping farce occurred only a few weeks before writing this article. The website Deviant art which has been in service for 22 years has open-source code released by Stable Diffusion) utilising all the members images to train the system without the members' consent. The community literally freaked out! The site has in recent days seen a mass exodus of talented artists who have been posting for years and the owners of Deviant art have now apologised and changed their policy on usage of their artists uploads.

For the people, but also for profit?

There is a technological race now to utilize the open-source technology and create new infrastructure and services that are valuable for people and industries. In my opinion the market will decide the winners the same way we have adopted new technology earlier. We will choose the platforms and the services that provide us with logical and easy to use user interfaces, speedy, snappy, reactive solutions, the feeling of fair value for money and the best output quality.

The industry has vast revolutionary potential and some of the best engineers and scientists are flocking to AI now. You might have noticed though, it has become very scientific. DALL·E 2 and Midjourney are partly founded as non-profit scientific research labs. By fronting the scientific aspect of their services, they are probably able to apply for grants and state funding. Lack of transparency might hide some of their for-profit minded activity. Clever use of lawyers and multiple subsidiaries in non-taxable geographic locations collect subscription fees and income. Price wise, with a few million users paying 15+ USD a month these companies have swiftly entered the big boys club with potentially generating 350—500 million dollars a year in subscription fees alone. That is a lot of cashflow and can be used as leverage to escalate scopes and services. It would be of significant moral, ethical, and creative advantage, as I see it, if these companies also came up with a novel way of paying all the contributing artists. All prompt queries from users that include reference to artists could as an example generate a micropayment to the artist. That could be a motivational factor and a more sustainable and future proof feature in my eyes. Let’s see what the future brings.



Refrences:

Crosslucid
AI and Art: A Dream to Some, A Nightmare to Others?