A.I.-Generated Art: Is This a Kodak Moment for Creatives?
Hardly any other topic dominates the current creative news feeds as much as A.I.-generated art. The latest iterations of these powerful image-generating A.I. systems just broke the internet. With tools like Midjourney, DALL-E 2 and Stable Diffusion, we have systems at our disposal that, at their core, have the power to completely transform creative work in general. They show us the power of generative deep learning and have been trained with hundreds of millions of images and captions. Based on these, they can create an original image of anything a human can describe with words.
With a single command (e.g., /imagine prompt) and a written specification of the image desired, artworks are created within seconds. The results can already be seen in marketing campaigns by Heinz and Gucci, and have been used to create NFT artworks based on the Bored Ape Yacht Club NFTs.
Tools like Stable Diffusion take it even further and let the users add personal references like a simple sketch that will help the A.I. compose the image.
By now the two most popular tools—Midjourney and DALL-E 2—differ mainly in the basic setting of the output filter. The images developed by Midjourney captivate with their surreal appearance, while DALL-E 2 produces way more photorealistic results. But both "styles" can be easily changed with appropriate commands. So the results can be adjusted accordingly by descriptions such as "paint it like a Picasso."
Click the images to enlarge:
The images above, from a DALL-E 2 blog post, are generated from the description "an astronaut riding a horse." One of the descriptions uses the prompt "as a pencil drawing" and the other "in photorealistic style."
While much of the current discourse on this A.I. art focuses on the result—how the A.I. interprets the textual input—it raises more and more controversy about how these tools have the potential to fundamentally revolutionize creative work.
If you look at the application area of image editing, for example, it means a simple text line like "paint a photorealistic donkey on a bike," produces a respectable result in seconds. A post-operator would surely need a couple of hours for that.
Or take time-consuming work like researching image databases. In the near future, this will not be necessary, when a simple prompt gives you the image you need in a blink of an eye. The need to brief an illustrator to draw your next campaign in a particular style? Just give the A.I. a reference picture and let the algorithms work their magic.
These examples show that a large part of the current value chain in the creative industry is facing enormous upheavals. And it does not stop here.
How will designers be trained in the future? Will we still need schools to educate? What about photographers or storyboard artists? And even top dogs like Adobe will have to catch up with their own tools.
This is only a glimpse of the disruptive changes that lie ahead. If we consider that these systems and their limitations are only at the beginning of their capabilities (most of them are in open beta), and add in the A.I. training capabilities of other databases, a tremendous part of our creative industry will have to deal with this transformation process.
Now that we witness the artificial intelligence being "only" trained with massive image databases, the question arises: What happens when this technology is combined with a chatbot and a Kindle library? Will these tools be able write books? If the underlying databases are the Netflix servers, will we be able to create our own personalized Hollywood blockbusters in the future at the push of a button? You can add any human-made creative output to that list—like architecture, music or fashion—and you will see there won't be any exception.
The question every individual in creative industry should ask themselves now is what role they will play as an art director, musician, author, graphic designer, illustrator, composer, VFX artist, copywriter, 3-D artist, lyricist or director in the future. The results we see definitely show us that these tools can replace humans in some way. Nevertheless, you will still need somebody to use the right commands to produce captivating work, which will lead us to a new understanding of creative jobs. It can already be seen that new types of working fields emerge.
In addition, I am sure the intrinsic idea of an artwork will increase in value and creative work will be opened up to everybody, even if you are not trained in a specific field. There is already a new mastery of creative work emerging with its own economic impact: the prompt engineer. On websites like promptbase.com creatives are selling their work, or rather the textual description of their work. So right now we are witnessing the moment when parts of the creative work are set apart from software skills or education. A moment when a broad range of creative fields are opened up to everybody and are truly being democratized.