Uncommon Courses is an occasional series from The Conversation U.S. highlighting unconventional approaches to teaching.

Art and Generative AI

I see many students viewing artificial intelligence as humanlike simply because it can write essays, do complex math or answer questions. AI can mimic human behavior but lacks meaningful engagement with the world. This disconnect inspired the course and was shaped by the ideas of 20th-century German philosopher Martin Heidegger. His work highlights how we are deeply connected and present in the world. We find meaning through action, care and relationships. Human creativity and mastery come from this intuitive connection with the world. Modern AI, by contrast, simulates intelligence by processing symbols and patterns without understanding or care.

In this course, we reject the illusion that machines fully master everything and put student expression first. In doing so, we value uncertainty, mistakes and imperfection as essential to the creative process.

This vision expands beyond the classroom. In the 2025-26 academic year, the course will include a new community-based learning collaboration with Atlanta's art communities. Local artists will co-teach with me to integrate artistic practice and AI.

The course builds on my 2018 class, Art and Geometry, which I co-taught with local artists. The course explored Picasso's cubism, which depicted reality as fractured from multiple perspectives; it also looked at Einstein's relativity, the idea that time and space are not absolute and distinct but part of the same fabric.

We begin with exploring the first mathematical model of a neuron, the perceptron. Then, we study the Hopfield network, which mimics how our brain can remember a song from just listening to a few notes by filling in the rest. Next, we look at Hinton's Boltzmann Machine, a generative model that can also imagine and create new, similar songs. Finally, we study today's deep neural networks and transformers, AI models that mimic how the brain learns to recognize images, speech or text. Transformers are especially well suited for understanding sentences and conversations, and they power technologies such as ChatGPT.

In addition to AI, we integrate artistic practice into the coursework. This approach broadens students' perspectives on science and engineering through the lens of an artist. The first offering of the course in spring 2025 was co-taught with Mark Leibert, an artist and professor of the practice at Georgia Tech. His expertise is in art, AI and digital technologies. He taught students fundamentals of various artistic media, including charcoal drawing and oil painting. Students used these principles to create art using AI ethically and creatively. They critically examined the source of training data and ensured that their work respects authorship and originality.

Students also learn to record brain activity using electroencephalography - EEG - headsets. Through AI models, they then learn to transform neural signals into music, images and storytelling. This work inspired performances where dancers improvised in response to AI-generated music.

AI entered our lives so rapidly that many people don't fully grasp how it works, why it works, when it fails or what its mission is.

In creating this course, the aim is to empower students by filling that gap. Whether they are new to AI or not, the goal is to make its inner algorithms clear, approachable and honest. We focus on what these tools actually do and how they can go wrong.

We place students and their creativity first. We reject the illusion of a perfect machine, but we provoke the AI algorithm to confuse and hallucinate, when it generates inaccurate or nonsensical responses. To do so, we deliberately use a small dataset, reduce the model size or limit training. It's in these flawed states of AI that students step in as conscious co-creators. The students are the missing algorithm that takes back control of the creative process. Their creations do not obey AI but reimagine it by the human hand. The artwork is rescued from automation.

Students learn to recognize AI's limitations and harness its failures to reclaim creative authorship. The artwork isn't generated by AI, but it's reimagined by students.

Students learn chatbot queries have an environmental cost because large AI models use a lot of power. They avoid unnecessary iterations when designing prompts or using AI. This helps reducing carbon emissions.

The course prepares students to think like artists. Through abstraction and imagination they gain the confidence to tackle the engineering challenges of the 21st century. These include protecting the environment, building resilient cities and improving health.

Students also realize that while AI has vast engineering and scientific applications, ethical implementation is crucial. Understanding the type and quality of training data that AI uses is essential. Without it, AI systems risk producing biased or flawed predictions.