Friday 30 December 2022

AI Art 2022



The AI Art (Artificial Intelligence Art) revolution accelerated in the second half of 2022.  After I got access to DALL-E 2 on August 18th, 2022 I went on a big AI rollercoaster ride, and it took over my life for the rest of the year:






I even exhibited my earlier AI creations of 2022 in a Virtual Reality gallery online -- one that visitors can walk through, even wearing a Meta Quest 2 VR headset:



Text-to-3D

Ugly but amazing!  By the end of 2022 a text-to-3D program came out -- POINT E.  I uploaded images to an online program -- Point-E text-to-3D Demo -- and it generated 3D files OBJ) that I could download.  Early in 2023 I generated and 3D printed some of these OBJ files:




AI Fantasies

I still want to exhibit my 3D figures in the Sculpture Garden in front of the Albuquerque Museum, even if I have to use AR (Augmented Reality) to pull it off.  AI made this fantasy more real, when DALL-E 2 introduced "outcropping" on August 31, 2022, and I expanded my AR snapshots:


My fake AR sculptures "outcropped"
 in front of the Art Vault Gallery
in Santa Fe




FIX MY FIGURE DRAWINGS!

Ultimately my main interest in the AI Art programs is to improve my figure drawings -- or better yet, generate something even more wonderful from them.  

For instance, AI applied R Crumb's style to one of my figure drawings -- the image on the right below -- and both improved my drawing and generated something wonderful and unexpected:




FAKE KRRRL Drawings



  • Some of these fake Krrrl generations can be seen online at the Open Art website -- @bee-svelte-61

AI has improved so much!  Back in 2020 I also generated fake drawings in my style, training a model with StyleGan2 in Runway ML, but the results were a lot more ambiguous, and not quite as impressive as those generated in the 2022 Photo Booth program:

COMPARISON:
Fake AI Krrrl drawings,
in 2022 and in 2020



AI Models for download
of Krrrl  Drawing Style

Photo Booth (Open Art), based on the first 20 drawings from my book:

StyleGan2 (Runway ML):

Justin Pinkney posted the "Long_Krrrl_Drawings" model above on his GitHub website -- awesome-pretrained-stylegan2 --  and filipecalegario (Filipe Calegario) of Brazil, linked to my model in his list of AI resources, under the "Artists and Artworks" heading:



Both Mark Hanslip (of the UK) and Benjamin Bertram (of Germany)  downloaded the "Long_Krrrl_Drawings" model and created something even better from it:

Mark Hanslip (ganslip on Instagramintegrated my model into one of his video commissions -- GANDERING 2-- for the Manchester Jazz Festival 2022:



Benjamin Bertram (ganwerk on Instagram) altered one of his portraits, posted it, and tagged me on Instagram:



It wasn't so long ago when I wanted the models to look like my drawings, since I figured that I was too old to improve my art skills.  Now I can lean on AI to improve my drawings!


UPDATE (Feb 6, 2023):  I just discovered that my AI model was used for a projection project in Kelowna, British Columbia, Canada --  Autolume Acedia -- my  model is an element of the second video.  More info at the bottom of this blog post.






AI ART Momentum
Accelerating Rapidly!!!


AI has come a long way since Holly Grimm gave us an AI presentation in 2018 at the Tuesday Night drawing group in Santa Fe, and exhibited her AI creations at Betterday Coffee in Santa Fe:




Holly wrote an AI Art program and posted it on GitHub -- cyclegan-keras-art-attrsShe exhibited both the poster of this program, and art pieces that she created with it, at Betterday Coffee.  She also was interviewed about her AI programs when interviewed on the Currents New Media podcast on January 7, 2021.

Holly also organized an AI Art exhibition in 2020 based on GPT3 -- AIKPHRASIS -- which was exhibited at the Färgfabriken gallery in Stockholm -- Sentient Machines? -- August 11 - 21, 2022.  I participated in this project:




AI Monthly Summaries 2022

AI improved so rapidly in the second half of 2022 that I had to write summary posts on my blog just to grasp what had happened:


August  -- I got access to DALL-E 2 on August 18th, 2022:
September  -- DALL-E 2 introduced "outcropping" on August 31, 2022:
October --  Deep Dream Generator introduced "Text-to-Image" on September 29, 2022:
November:

         -- Midjourney upgraded to "Version 4" on November 5, 2022:
         -- Stable Diffusion 2.0 released on November 23, 2022:
           --ChatGPT became available to the public on November 30, 2022:
December --  POINT E "text-to-3D" demo programs become available in late December:






CONCLUSION

Who needs to draw anymore? How about voice-to-image:

model drawings created by Midjourney,
no pencils involved


More critical though, may be that AI is pushing art into a few particular narrow directions, probably depending on how the underlying models were trained.

I wrote on November 21, 2022:

"There is a certain default Barbie factor to Midjourney. Ambiguous prompts seem to converge on a female portrait, and one that is more of a single idealized portrait that looks like Barbie's face. When the result gives a full body, it tends to be one with Barbie proportions."

 November 21, 2022




Moreover dissing AI Art became the major past time on the Internet in the second half of 2022.  The most interesting story was when Ben Moran of Vietnam was banned by Reddit for posting AI Art -- for an image he created in Photoshop, not with an AI program.

When is the work of Polish illustrator Greg Rutkowski going to be banished from Reddit and the Internet?  AI artists are constantly invoking his style -- such that his style might become synonymous with the style of AI art.  

Will the anti-AI algorithm ultimately ban Greg Rutkowski?


I'm not worried.


...and ChatGPT isn't helping either:




No comments:

Post a Comment

Note: only a member of this blog may post a comment.