In the beginning of the 2020s, in response to advances in artificial intelligence and deep learning, softwares like GPT-3, Dall·E, Midjourney, and Disco Diffusion have enabled the generation of images based on textual instructions. Although the results are quite prototypical since they merely imitate based upon existing data, this has sparked a series of polarised media debates on issues regarding the replacement of humans by machine learning. In doing so, the critical question is passed over: what are the actual pitfalls when considering the current and potential implications of machine learning for design practice?