[photochemsyn]

People across the government-corporate spectrum are now fully aware that if they train 'self-learning algorithms' on biased data sets then those algorithms will adopt those biases. Such people seem to believe they can then get away with blaming the algorithm for the outcome while avoiding any legal responsibility themselves.

On the positive side more and more people are generally aware of how this works. Criminal investigators need to look at the people who selected the datasets that were used to train the algorithms, they're the ones who bear responsibilty.

[masklinn]

> People across the government-corporate spectrum are now fully aware that if they train 'self-learning algorithms' on biased data sets then those algorithms will adopt those biases.

"Bias laundering" is how it's usually called. Take biased training data, put it in "the algorithm", and presto it's all legal because it comes from "the algorithm".