• Home
• /
• Lifestyle
• /
• Arts

Trending: UAE artist behind viral 'Satwa Superheroes' decodes AI-generated art and the ethical concerns it raises

Yohan Wadia on why AI art made through platforms such as Lensa AI and Midjourney is drawing an equal mix of praise and criticism online and its implications on copyright infringement

Over the last few weeks, fantastical AI-driven portraits and selfies have been all over social media, with excited users hopping on the trend across the globe. While scrolling through the Instagram feeds, one is bombarded by photos of friends and family members turning themselves into mythical characters — superheroes, medieval princesses, historical figures, movie characters — you name it, it’s there.

The enabler behind this viral trend, however, is not the stroke of a paintbrush but a click of a finger. Users have been flocking to platforms such as Lensa AI and Midjourney that allow them to create AI-generated artwork and ‘magic avatars’ of themselves, through inputting prompts and suggestions.

Amongst the flurry of AI-selfies, a UAE-based artist also hopped on the bandwagon recently but with his own unique twist. An illustrator and creative director by profession, Yohan Wadia has reimagined blue-collar workers from Dubai’s old neighbourhood of Al Satwa as superheroes from the Marvel and DC universe. From Superman on a streetside bench to Spiderman walking across with grocery bags, the artist has created AI-generated artworks to celebrate the unsung heroes of Dubai’s commercial hub, who according to him, make the city what it is.

However, even though his ‘Heroes of Satwa’ series took social media by storm, Wadia, along with other digital artists, has also been on the receiving end of criticism related to AI-generated art. Among other ethical concerns raised by these apps, artists using these platforms have been accused of copyright infringement and oversimplifying the process of creating art.

We spoke to Wadia, who has created part 2 of the viral ‘Heroes of Satwa’ series exclusively for Khaleej Times, to take us through the steps required to create AI-art, while addressing pertinent ethical concerns being raised against this new technique.

Q) How would you describe AI-generated art?

AI art is generated with the help of machine learning. It is developed with the help of computing power that scours the internet, learns patterns, styles, and about any other attribute pertaining to art, and recreates visuals based on the learning with the help of user-input prompts.

Q) What are some of the apps/platforms one can use to create AI art?

The best in the game are Midjourney, Stable Diffusion, and Dall-E. I personally use Midjourney for my art.

Q) Can you briefly explain the steps required to create AI art?

While most people think AI art is as simple as pushing buttons and getting results, it’s unfortunately not. The very first thing one needs to do is visualise with as many details as possible and then punch in the description of the image they have in their mind, to be able to generate something as close to their imagination as possible. The AI then generates some variations based on your prompts and you can go about further fine-tuning the prompts depending on how the AI produces the art until you are able to perfect it.

Q) You need to input data sets to create AI art. Are these data sets derived from copyrighted content?

The AI learns from any and every source on the internet, hence, it is able to reproduce derivatives of content that may be copyrighted, although it doesn’t replicate anything to the absolute detail of falling into copyright infringement. A publication recently reported that using copyrighted material to train AI models might fall under fair use laws.

Q) A criticism of an AI art-generating app online mentioned: ‘This app is making a profit on a mountain of stolen, uncredited and uncompensated art’. To what extent is this true?

Nothing we ever think of or create is original. Every invention or every concept is a derivative of an experience or a specific encounter in the past. While AI learns from patterns, techniques and imagery, it creates visuals that are completely original and differentiated from one another. You’d never find two of the same in AI-generated imagery.

I am an illustrator and creative director who has spent the last 14 years of my life designing and as an artist, I have constantly thrived on inspiration. But I can guarantee you that it is always something or the other that inspires us to come up with original pieces of content. It’s even possible that AI may have scoured through some of my work to generate artwork if someone had prompted something that matches my style. However, that is the way art has always worked. Unless something is blatantly stolen and replicated, art inspires art!

Q) What are some other criticisms that artists have received for AI art?

The most common one is ‘AI art is not art’. Well, in that case, digital paintings are not paintings and digital camera photography is not photography. It is unfortunate that most criticism primarily arises out of insecurity and fear. Fear of being replaced by something superior and faster than us humans. There has always been resistance and rejection at the inception of every invention and AI art is no exception. I myself was taken aback looking at the possibilities and my first reaction was, “Well, looks like we’re losing our jobs soon”. But upon thinking, it is only a tool that facilitates execution. Execution has been overrated for the longest time in history. It involves a lot of painstaking work, there’s no doubt about it, but the time has come when technology is making it easier for the layman to be able to execute something spectacular with a simple thought! Isn’t that amazing? It is what the mobile phone did to photography and what the computer did for several other professions, including graphic design.

Q) The AI selfies created through Lensa AI and Stable Diffusion have also led to privacy concerns amongst users…

One of the main ethical concerns of AI-generated work is around companies misusing user data. However, an article I read online states that the images used to train the AI are purged once the learning is completed. Lensa AI made a public statement on Twitter mentioning that the photos and the ‘associated model’ are permanently erased from their servers as soon as the avatars are generated. Hopefully, there will be more clarity around these concerns going forward.

somya@khaleejtimes.com