Lack of large parallel corpora or dictionaries for most language pairs and modern data-hungry models make the need for unsupervised translation pressing. In this talk, Sebastian introduces us to recent developments in the field.