Democratizing AI and Attacking Algorithmic Bias


Date & Time

Tuesday, 26 Feb

13:00 - 14:00


Hall 4 - Auditorium 5


The conventional wisdom is that artificial intelligence should be the great equaliser when it comes to bias in decision-making. Decisions become all about the data, cold calculations unhindered by human bias or prejudices. Yet already, this perception has been revealed as a fallacy, with numerous examples of AI fails resulting from skewed input data. Without stringent controls, AI is liable to stereotype and amplify sexist and racist biases from the real world, reinforcing existing social inequalities for people who are already vulnerable. The key to avoiding algorithmic bias is in striving to avoid inadvertent discrimination in the data that is provided to the AI in the first place. Compounding this challenge, AI intellectual property is concentrated amongst just a few global players who invest billions of dollars in R&D to maintain their position of leadership. While its impact is ubiquitous, control is concentrated. Unfortunately, a centralized approach to AI opens the door for potential abuse of power. Democratizing AI means distributing control and improving the AI black box, detecting hidden bias and providing explanations for the automated decisions being made. This eye-opening session will provide an informed perspective on how we can work together to eliminate algorithmic bias and democratize AI.

Download the official #MWC19 event app