© Pint of Science, 2020. All rights reserved.
From self-driving cars to online shopping advisors, more and more of our life is being guided by computers who have taught themselves to assist humans with tasks. But is such assistance always helpful? Join Miranda Mowbray, who will speak about how such computers are used to help sentence criminals in the justice system, and their limitations in the application. Also speaking is Kit Fotheringham, who will explain the risks of these technologies entering the public sector administration on a large scale, and how these risks can be avoided by engaging social researchers with public bodies and th...
Predictive Analytics: You Can't Have it All
Miranda Mowbray (Lecturer in Computer Science )
This talk will be about limits to predictive analytics. The main application I'll talk about is to algorithms advising on the sentencing of convicted criminals, but the limits are true for predictive algorithms more generally. I'll show that some of the properties that people would like predictive analyses to satisfy are inconsistent, and argue that the choice of which properties should be satisfied is a policy decision that shouldn't be left to data analysts alone. Just for fun, my talk will include 79 and a half cats.
The Learned Machines are Coming!
Kit Fotheringham (PhD Student)
Bristol is home to some glistening trophies of civil engineering but also bears the scars of the Industrial Revolution. The common assumption is that technological innovations usually represent progress. Most of the innovations occurring in the 21st century have been described as contributing to the 4th Industrial Revolution, namely the Information Revolution. Artificial intelligence and machine learning are the latest iteration of these technologies. This talk will explore risks to democracy and personal liberty when AI technologies enter public sector administration on a large scale.