AI-Generated Art Reflections

I’m exploring AI-generated art techniques using code originated by @rivershavewings, and @somewheresy and further developed by a large and growing online community of artists and developers. My goal is to create sets of interesting and provocative images and author numerous short animated films.

This page is intended to describe my process. I am focusing on exploring AI-generated art with a Chicano theme. I’ll continue to document my research here in this post and upload some of the results.

Because of the ease of creating imagery with this toolset the foremost issue is becoming how the imagery is depicted and sourced, in other words, Design Style. Currently, a heavy influx of imagery leverages pop culture material from game design, popular novels and authors, television, and cinema. There is no true vision that is emerging other than derivative content.


Though local installs on Linux workstations are often the choice of power users. The data generated in this technique is largely done with Google Collab notebooks. This technique allows the user to remotely obtain a dedicated Linux computer to generate the artwork. I’m currently using Disco Diffusion and JAX2.7 though there are many other powerful notebooks out there. There is a strong and supportive community on Discord and Twitter. To find them on Twitter just do a search for #aiart or #aiartcommunity.


ARTICLE-CLIP: Connecting Text and Images

We’re introducing a neural network called CLIP which efficiently learns visual concepts from natural language supervision. CLIP can be applied to any visual classification benchmark by simply providing the names of the visual categories to be recognized, similar to the “zero-shot” capabilities of GPT-2 and GPT-3. https://openai.com/blog/clip/


AI in Art Newsletters
http://newsletter.unlimiteddreamco.xyz/
https://www.getrevue.co/profile/VITO_AI
https://us15.campaign-archive.com/?u=c7e080421931e2a646364e3ef&id=9efba7e932
https://makeaiart.substack.com/p/ai-artwork-newsletter-new-classes?s=r


The most useful reference site for tools and techniques is maintained by @pharmapsychotic: https://pharmapsychotic.com/tools.html.