With pretrained language models and generative AI systems, machine learning is being embedded into the social fabric at an ever-increasing pace. However, how such socio-technical objects come to engage with society remains less understood. In this talk, I will explore the phrenological heritage of machine learning, how the distributional hypothesis affords creating fictions of humanity, and how we in turn, humanise that which imitates us. In closing, I will discuss the broader social implications of the current generation of machine learning artefacts in society, and reflect on the roles we hold as researchers and practitioners