Keynote: Surveillance, Power and Accountable AI Systems: Can We Craft an AI Future that Works for Everyone?

Summary form only given. The complete presentation was not made available for publication as part of the conference proceedings. Tech leaders and even those of us building software systems for governing our world like to imagine the ways in which automated systems might enable a better future - curing diseases, lifting people from poverty, saving animals from extinction. However, there is plenty of evidence that instead platforms and algorithms are contributing to directly to huge societal problems like spreading of misinformation, political polarization, rising inequality and more. I will talk about the role of microtargeting of individuals based on digital surveillance, machine learning based on historical data and black box algorithms based on a lack of explainability and unaccountability. If we want automated decision making that protects the rights of individuals and healthy societies, that reduces bias rather than amplifying it, we are going to need to engage in real discussions of the tradeoffs with stakeholders ‚Äì not just empty promises that “this is going to be great for everyone”. We need to disrupt the current levels of digital surveillance, shift control over surveillance data, require explanations of automated decisions, enable aggressive third-party testing, and create incentives for iterative improvement in systems where those being impacted by the automated systems are not the customers.