This interactive installation paper describes [self.], an open source art installation where the people interacting with it determine its auditory and visual vocabulary. When the system starts, it knows nothing since the authors have decided that it should be without any kind of bias. However, the robot is equipped with the ability to learn and be creative with what it has internalized. In order to achieve this behaviour, biologically inspired models are implemented. The robot itself is made up of a moving head, mounted with a camera, projector, microphone and speaker. As an art installation, it has a clear robotic visual appearance, although it is designed to demonstrate life-like behaviour. This is done by making the system start in a “tabula rasa” state, forming categories and concepts as it learns through interaction. This is achieved by linking sounds, faces, video and their corresponding temporal information to form novel sentences. The robot also projects an association between sound and image; this is achieved using neural networks. This provides a visual and immediate way of seeing how the internal representations actually learn a certain concept.