How does an AI system understand image and text? How can we navigate ourselves in such systems that play around text and image with assumptions, biases, and prejudices? With some simple programming, we could gain insight into how the black box operates. The course is an introduction to how we, as creative workers, can utilize generative AI while maintaining a critical standpoint. All image-generating models are based on a system that translates between language and images, an interesting concept where visual and linguistic meaning intersect. Human vision is a deeply contextual process, drawing on personal history, cultural frameworks, and embodied experience. Yet Machine vision operates on statistical analysis of vast datasets, reducing image complexity to mathematical relationships. In practice, we will install small-scale text-image models that could operate on our own computer without depending on cloud-based services. Explore the mechanisms of these vision AI through hands-on programming practices. Finding where the biases came from and how we could emphasize them by artistic means. Bringing the possibility of developing visual-based installation or moving image practices. We will read and discuss ideas from various perspectives and references, engaging with diverse voices, and contextualizing our practice within broader conversations of critical AI. Each technical exploration will be paired with critical reflection and artistic references on the implications and assumptions embedded in these systems. * Course Introduction & Environment Setup * Generative AI Mechanisms * Terminal and small-scale Generative AI * Python Programming & and local AI (Ollama) * Image Generation Workshop with Artist Designer Fabian Mosele * Text-Image Relations (CLIP) and Biases * Vision-Language Model (LLaVA) * Operational Images |
