Interview on the Use of AI in Creative Professions
Andre Weiers
Freelance Senior 3D Artist & Animator | Ich belebe Eure Inhalte mit realistischer CGI & wirkungsvoller Produktvisualisierung
Question 1.
When and how was your first reaction when you learned about text-to-image generators?
?
First Encounter with Dall-E: Skepticism Meets Curiosity
I was absolutely thrilled when I first heard about text-to-image generators. Initially, I was somewhat skeptical upon my first encounter with Dall-E, as the early results were merely primitive attempts and showed little originality, akin to most images created by so-called 'AI artists' using these generators.
?
Midjourney: Stunning Image Generation Despite Challenges
However, my perspective shifted when I signed up for Midjourney. To simply describe an idea and see it turn into an image is truly breathtaking! Naturally, I realized rather quickly that it often takes several attempts before the outcome meets one's vision. Sometimes, though, unexpected results are produced that stimulate the imagination and invite further creative thinking
?
Documentation Gaps: The Enigma of AI Tags
One aspect that stood out to me was the insufficient documentation regarding the tags learned by the AI systems. These focused primarily on elements such as camera, lenses, perspectives, etc. It appeared that the qualification of the images had been somewhat overlooked.
?
Surprising AI Interpretations:
The 'Female' Butterfly Phenomenon
An intriguing instance was when the term "A Female" in Stable Diffusion conjured an image of a female butterfly. This was because "Female" seemed to be a learned "tag", derived from the biological photography of thousands of butterflies. It emphasizes the fact that these systems are classic neural networks that rely on probabilistic functions to generate results.