Gender bias in AI: building fairer algorithms

Machine learning systems are what they eat. Since the training data for NLP problems is human language, these models have a tendency to perpetuate our own biases. How can we build better algorithms?