The new “Algorithmic Accountability Act” that has been put to congress forces large companies to assess their machine learning algorithms for bias or discrimination. The bill categorizes “large” as any company making more than $50 million a year or holding data of over 100 million users. An increasing amount of large companies are employing AI powered tools such as ad targeting and facial recognition but while the technology is young they continue to show technical flaws. Amazon recently scrapped an AI recruiting tool that was found to show bias against women and M.I.T.’s media lab illuminated how algorithms have greater error in identifying darker skin tones.
