In recent months, the creative industry has discovered generative artificial intelligence (AI) tools such as DALL-E 2, Midjourney or Stable Diffusion and has been able to see first-hand the amazing potential housed in the bowels of artificial neural networks. Whether we are talking about photographs, illustrations, logos or icons, the creativity born from the womb of AI is so astonishingly good that almost anyone can reach («ipso facto» and with little effort) the status of artist, illustrator or graphic designer.
However, generative AI is also subject to a strong polarisation . Some see programmes like Midjourney as perverse because they have been pre-trained with huge amounts of data provided by designers, photographers and artists (so they are by no means immune to plagiarism). In addition, these tools (still rudimentary) are not yet sufficiently flexible in the eyes of many. Yet many others are convinced that AI is laying the foundations for a new era in the creative industry.
Ultimately, programs like DALL-E 2 and Midjourney could prove to be absolutely croatia number data indispensable in this field of activity , especially in the early stages of creative projects, emphasizes Michael Katzlberger in an article for Horizont .
Ideas, sketches and storyboards can suddenly be generated at lightning speed and easily implemented in presentations, campaigns and pitches . This is a first-rate competitive advantage for the creative industry.
“Stable Diffusion” (SD), the technology behind tools like DALL-E and Midjourney, first debuted on the market in 2022. And compared to GAN (Generative Adversarial Network), SD is a much more democratic technology and perfectly accessible through an “open source” model.
SD is a deep learning model that is primarily used to generate images based on textual descriptions. However, this technology can also be used for techniques such as inpainting (restoring destroyed or lost parts of an image) or outpainting (continuing images beyond their original boundaries).
It is worth noting, however, that truly excellent AI-powered images are only possible with highly polished texts that require a lot of time and patience. Perhaps that is why one of the most in-demand jobs in the creative industry in the coming years could be that of a “prompt designer.” After all, the more detailed the text descriptions that DALL-E and company are fed, the better and more exciting the results will be. This is good news for those in the creative industry who boast a very broad vocabulary, and also for those with advanced knowledge in the field of photography and video (to give AI clues about the most suitable lens and lighting when illuminating images, for example).
Although SD-based programs boast phenomenal “superpowers,” they also struggle with problems similar to those faced by the older GAN technology. For example, these programs often produce images of very low resolution. It is also sometimes difficult to realistically depict the limbs of humans and other living beings, which is why DALL-E 2 and company feature images of people with more than ten fingers, two tongues, and up to four rows of teeth.
In the future, AI could make other inventions born from the hand of man pale in comparison.
Image generation is, however, only one of the possibilities of SD models , which with appropriate textual descriptions can also be used for the generation of videos and 3D models.
In addition, OpenAI's most advanced artificial intelligence to date could be hatched in 2023: GPT-4. This AI will be equipped with multimodal skills, so it will be able to not only write texts but also generate images and compose music.
Creating images through textual descriptions is merely one piece of the puzzle to complete the quest for the Holy Grail, the so-called strong artificial intelligence or AGI , which could become a reality in the next few decades and free AI from the silos in which it is currently anchored.
This technology would combine the progress made by artificial intelligence over the past decades and would inevitably eclipse other human inventions . For the first time in human history, children's role models could be artificial intelligences and not their parents or teachers.
Although it is often covered in a very disturbing patina, artificial intelligence is destined to play a leading role in the future of the creative industry because it frees humans from their genetic constraints and expands creativity by blowing up all boundaries, concludes Katzlberger.