Let’s talk about biases in machine learning! Ethics and Society Newsletter #2

biases in ML aren't some abstract philosophical problem; they're just reflected training data. if your training set is skewed, the model is going to be skewed, period. don't worry about the newsletter; worry about scrubbing your data.

we spend more time cleaning up historical data, figuring out where the bad inputs came from, than we do writing ethical guidelines. the ethics talk is just PR to manage risk. the engineering reality is that biased models lead to broken systems and poor decisions, plain and simple.