I primarily use midjourney’s generative image AI for my creations by prompting specific keywords to influence the aesthetics of the resulting images, e.g., « warm natural colors, » « 1970s, » etc. A stunning example of what you can get out is the example in this brush that draws a furious brush, in this picture …

The images generated by AI models like Midjourney, Dall-E, Stable Diffusion, etc., are astounding. But all of this comes at a cost…

too long to read? Scroll down and get the best pictures out of my AI promting…

As I walk along with AI, I get more and more aware of the negative aspects of these tools: generating images with AI is all but neutral, neither in terms of carbon footprint nor ethics.

I’ve pondered many questions during my work… questions about computer resource consumption, copyright violations, the future of creative professions…

I don’t have all the answers – but regarding energy resources: the carbon footprint is terribly negative, and I’m stopping for that reason primarily – even though I’ve paid several hundred dollars for my sustained use of Midjourney (it’s not free… let’s hope it’s used for good), I wouldn’t want to generate more images in the future for ecological reasons. It’s easy to say in hindsight… I’ve built up my stock for my ai-nigma project: over 6500 images generated by me, including the samples below.

In September 2023, I canceled my Midjourney subscription mainly for this reason. And also because I have concerns about the unrestrained use of these tools in envisioning the future: generative learning models are already feeding themselves with their own results, creating uncertainties and errors. Inevitably. And all of this doesn’t smell well. Recent studies support this view. (https://arxiv.org/abs/2307.01850) – Kind of a self-feeding SISO (Shit-In-Shit-Out) loop building up and filling the space with slowly growing disorder and uncertainty ? We’ll see.

Regarding copyright, I remain skeptical. The illustration image here shows a brush painting a disgruntled brush-face… Did the face exist in that expression? With brush bristle hair? Highly unlikely! Then there’s the question of the initial data collection for the « training » of these AIs: certainly not respectful of copyright: the more, the better.

To conclude, here’s my opinion on creative professions: YES, those who already rely solely on plagiarism, youtube tutorials and copy/paste mentality should be worried: artificial intelligence can do plagiarism just as well or even better, at a lower cost. Let’s comfort others: generative image AI – and I dare say AI more broadly – cannot be creative in the noble sense of the term; it can only reproduce… through its blending capability, meaning it can create a credible fusion of different source images. You may argue: it can also quickly generate less artistic content; you can for example ask it to create an image of a coffee served on a table in front of a sunset – this is where creative professions will be in danger: paying a photographer will be more expensive than generating an image.

For me, the artistic appeal comes from its ability to stunningly merge the existing. It can, based on what already exists, become inspiring through its immense combinatorics power. That’s how I’ve used it. The example of the lemon jewelry below is so convincing.

Create beautiful things, of your own!

… and here are the most beautiful images from my collaborative work below, considering all my remarks… I don’t claim copyright on these images – the debate on this subject is currently ongoing – but I would still appreciate you to reference my project if you use them in other places. All images on my website contain a visible « imavia » stamp in the bottom right corner: I may provide the original one on request. Some of them are also « ownership-acquirable » via NFTs.

That said, indeed, after exploring the possibilities of generative image AI, I intend to return to old friends: acrylic, watercolor, … and classical generative art (programmed, robotic,… but without or with less AI!)