While machine learning algorithms are driving lots of large-scale technology—from recognizing human speech to self-driving cars—human bias creeps into the code in unsettling ways. The rules and the input data that control algorithms can make them discriminate further against women, minorities, or other groups if such a bias seems to already exist in the data.
A new group of leaders in science, business, and tech is looking to change that. Based in the United Kingdom, Women Leading in AI propose getting ahead of algorithmic bias by creating a set of standards to encourage thoughtful governance of artificial intelligence systems and algorithms.
Although the 10 regulatory recommendations from the group are focused on the UK, they seem quite reasonable as a conversation starter for anyone working in the field of AI. The points cover not only establishing a system of oversight and compliance for algorithmic development, but also encouraging transparency and skills development in the burgeoning AI workforce. The recommendations also include the idea of encouraging good conduct by reducing liability for "companies that have obtained a Certificate of Fairness to foster innovation and competitiveness." --Yazmin Villegas