Sep 12, 2022
here's a lot of buzz about so called AI generated art at the moment using programs like Dall-E, Neural blender, Dream studio lite and so on. People use text prompts, i.e. “flying clockwork fish balloon” and the AI generates “art” based upon that. It usually does this by using actual art and photography created by real people, and visually mixing them to create a new image. Some results are much better than others.
Topics and Show Notes
The main issue with the approach is that the AI like all AI is dumb: it has no awareness, no understanding, it can't think, it can only serve up the most superficial interpretation of the prompt text. 100% of the “meaning” or “art” a viewer sees in that image is their own imagining, this in itself is never a good indication of art. But this also means that the AI has no capacity to understand the visual language it's speaking so it just provides you with a poor facsimile of art, which is all it can ever do due to inherent limitations.
For example, ask it to create a pirate image based on certain parameters. All it will be able to do is create a pastiche of already created images on the subject, like a bad deviant art image by a poor artist. A good human artist would think about the subject in terms of the ethnicity of the pirates, the type of vessel they use, how ratty or decorated their clothes are due to their time at sea, the body hair, tattoos, piercings and injuries on the pirates, the time of day and where the light comes from, the style of clothes and weapons due to the period, geography, and culture of the pirates etc. All that means something and will have a significant influence on the image (you can apply this to any other scenario too), while AI can only produce generic pretend art.
Unfortunately people don't understand these significant limitations and expect AI to do more than it can, (in all fields, not just art). AI is a useful tool when used correctly and when you properly understand its limitations, but when you don't it creates havoc. Which is why AI assisted systems tend to be so shit- like auto-correct spellcheckers, Google search algorithms, Facebook and Instagram image checkers, YouTube copyright checkers, “smart” chat-bots that big companies use as help systems etc. If we have people mistakenly thinking AI can replace artists or graphic designers, then we're in for a tsunami of boring, generic, meaningless design and the world will be a blander place because of it.
Try not to be fooled by the “idea” of art, when you see an AI generated image remember there was no thought behind it, no meaning to the placement of figures, style of clothes, body posture, mistakes etc, and the stylistic choices are simply stolen from human artists. It's a system that specialises in creating pointless parody pastiches.
This week Gunwallace has given us a theme to Brittany Hyperstyle - Bouncy, pop bubblegum bubbles of light and joy floating, rising and falling in sweet smelling convection currents. This sound is pink and powder blue with a sheen of iridescence. Synth electronica that’s fun, light and happy!
Topics and shownotes
Forum discussion that inspired the topic - https://www.theduckwebcomics.com/forum/topic/179155/?page=1#latest/
Dream studio lite - https://beta.dreamstudio.ai/dream/
Vast as the Night - https://www.theduckwebcomics.com/news/2022/sep/06/featured-comic-vast-as-the-night/
Brittany Hyperstyle - https://www.theduckwebcomics.com/Brittany_Hyperstyle/ - by Shampoo Venom, rated T.
Special thanks to:
Gunwallace - http://www.virtuallycomics.com
Ozoneocean - https://www.theduckwebcomics.com/user/ozoneocean
Banes - https://www.theduckwebcomics.com/user/Banes/
Tantz Aerine - https://www.theduckwebcomics.com/user/Tantz_Aerine/
PitFace - https://www.theduckwebcomics.com/user/PIT_FACE/
Become a subscriber on the $5 level and up to see our weekly Patreon video and get our advertising perks!
Even at $1 you get your name with a link on the front page and a mention in the weekend newsposts!
Join us on Discord - https://discordapp.com/invite/7NpJ8GS