"If you don't know what a model was trained on, who it was trained for, and what it fails at — you don't know the model. You only know its output. And output without context is the definition of a stochastic parrot."
Chief Ethics Scientist at Hugging Face who co-invented Model Cards, now the industry standard for documenting AI models' capabilities, limitations, and risks. Fired from Google AI alongside Timnit Gebru, she has continued building the infrastructure of responsible AI practice — documentation standards as coordination architecture.