Microsoft's Azure AI platform will be updated to feature anti-bias software. The scientists that that develop AI tools often do so with their own biases, leading to technology that increases discrimination. In an effort to improve upon these practices, Microsoft's Azure platform will be updated to feature its own Fairlearn toolkit, which helps companies building AI software test for their own biases.
The toolkit, which will be available in July, enables data scientists and developers to see not only what biases slipped through the cracks, but how those biases could impact the lives of the user. For example, the toolkit was used in testing on EY's automated loan algorithm, ultimately revealing the gender bias within the system. According to the toolkit's results, EY's model was 15.3% more likely to give men loans versus women.
Image Credit: Eduardo MunozAlvarez/VIEWpress via Getty