The problem I see is that AI systems can also perpetuate existing biases by using feedback loops. For example, if an AI system is used to recommend job candidates, and it recommends more men than women, the system may learn to recommend more men in the future, perpetuating the bias.