A machine learning system that can (1) parse real time human spoken language and reliably detect any hint of bias which can then (2) output an unbiased version of that input.

Some use cases would be to deploy this in government (e.g. courts, police force etc) and for the betterment of humanity in making humans aware of their unconscious biases in decision that have far reaching consequences if left unchecked.