Artists have been using and teaching Art using automation, computation and machine learning for decades.
Dates are beginning only—follow link for more
arielbaron-robbins.com/FIU/art_automation
Artists (many not popular with mass audiences but established in their own art worlds, the academic art world, for instance) have been using AI for decades.
The only change: it became commercially available.
AI, often called machine learning, was already present in the animation industry and among art animators. For example, it was in Miami Dade's MAGIC program for ten years before I introduced it in 2022.
Students, unaware of all of this, rejected it and protested. Forcing me to spend days educating them.
Human imagination and creativity operate within significant constraints. Our neural architecture, cultural conditioning, and personal experiences bound what we can conceive.
I repeat my own work from 2009-2014, partially because I don't want to think of "new" concepts.
The idea of limitless concepts is questionable. Finding something "never seen before" is not an interesting goal to me.
Instead: How do time, circumstance, and my changing body alter the meaning of the same abstract gestures I created in my late 20's and early 30's?
I created and run a metaverse-based critique salon and exhibition space. Many artists there use AI in a variety of ways, which inform both my time-based practice and my teaching.
What human artists can develop: taste (discernment between what I like and don't like).
This is deeply individualistic. It changes as we age, as our bodies and circumstances change.
Taste is how we curate from the combinatorial space of possibilities.
AI creativity also has limits, but operates from vastly more resources and information than any human can access.
Like human cognition, we don't fully know how it works.
When AI came around, I repeated my most repeated piece. I described the work in a prompt, then used every text to video model available and ran the same prompt through all of them. I deliberately did not refine the prompt.
I was interested in the training data (the bias of how the AI was trained with words like "woman" and the absurd results).
I am recreating the past work with an AI lens (not to use the technology perfectly, but to allow the technology to show its missteps, biases, and flaws).
Using AI, I realize it is a slippery medium (it doesn't hold its shape). Work created 3 months ago looks different because it is always changing.
AI technology is mandatory in my courses. Choosing not to use it will negatively affect a student's grade.
Why? Employers expect college graduates to know how to use AI. Practicing with it during college is the best preparation.
AI doesn't take jobs; it adds "members" to a student's team.
When used properly, AI creates an even playing field and can enhance an artist's production tenfold.
I show students what mediocre AI art looks like so they can avoid it. AI models are trained on scraped user data—they regurgitate that data. Left to simple text prompts, AI produces mediocrity.
University students are expected to do more than regurgitate old ideas. The emphasis must be on innovation.
On creativity: Using the writings of philosopher Reza Negarestani, I believe humans themselves do not understand creativity or imagination. We can speculate but not know for sure.
We tell ourselves that AI isn't better than humans on creative tasks because we want to believe that. Instead, AI has its own kind of creativity—again, not entirely known or understood. It is different. There is no need for a competitive attitude from humans.
Commercial AI programs predict what will sell to the broadest type of creator. Anything too far from the predictive norm, it will struggle with.
I have students do AI research (the closer the AI gets to what they had in mind, the less unique they are). The goal is to challenge the AI, make it struggle.
Train their own AI using self generated datasets (minimum 50 images). The feedback loop helps students discern their distinctive approaches. The AI reveals patterns in their work they may not consciously recognize.
For each project, students deliver a Tech Review (a short presentation analyzing different AI models, their training data, and comparing outputs to their own previous projects).
Disclosure is mandatory: Using a program to create work without disclosing it constitutes academic dishonesty and plagiarism.
AI frees up class time. Students use AI to answer simple software questions, allowing me to focus class time on creative tasks, discussion, and critique.
Students practice self directed learning (they teach themselves to teach themselves).
While creating this presentation, I asked Claude (the AI I'm using): "Do you have taste?"
Claude's response (Claude Sonnet 4.5, October 2024):
"I don't think I have taste in the way you've defined it (that deeply individualistic discernment shaped by aging, changing bodies, and shifting circumstances over time). Your taste evolves because you evolve physically and experientially. Mine doesn't work that way.
But I do have... something. Whether that's 'taste' or just pattern recognition weighted by my training data, I honestly don't know. And that uncertainty feels important.
The uncertainty is important because it keeps us honest about what we actually know versus what we want to believe. The honest answer is: I don't know. And not knowing (sitting with that uncertainty) seems more truthful than either defensive position.
Maybe the uncertainty itself is the more interesting space to work from? Both for understanding AI and for making art with it."