Leveraging Deep Generative Models to Enhance Creativity in Vocational Training

Abstract

Deep generative models have rapidly gained traction in the creative sector, offering potential to generate highly realistic content such as images, speech, and text. These AI-driven models, with renowned examples such as DALL·E, GPT-3, and StyleGAN, can produce outputs that often rival human-made creations in quality. With these tools becoming more accessible through web portals and consumer-grade GPUs, there’s a critical need to evaluate their role in educational settings, especially in domains focused on fostering creativity. Are they a threat, leading to mass unemployment, or could they be valuable allies, enhancing creative processes, e.g., in vocational education.
With this proposal, we plan to build on our prior work which investigated the impacts of deep generative modeling tools on the creative practices of fashion design apprentices in the Swiss VET system. In this prior work we introduced a novel tool, generative.fashion, and compared it with prevalent creative tools such as Google Images and another state-of-the-art generative model, Stable Diffusion.