People seem to be trying to answer the practical question "How do we build a non-racist model?" the answer to whi h depends entirely on the philosophical question "what does it mean for a model to be racist?" the answer to which no one can seem to agree on.