Safe, secure and robust AI for safety-critical systems

Professor Des Higham is Guest Editor for a theme issue of the Royal Society's Philosophical Transactions A, co-edited with Ajitha Rajan from the School of Informatics.

Handwritten text that appears to read, "A robot may not injure a human being"

The writing above was interpreted by a state-of-the-art AI system as "a robot may now injure a human being". How did this happen? 

Des Higham and Lucas Beerens exploited known vulnerabilities to change the pixels in the original digital scan in a way that causes the system to malfunction. 

Learn more in Des' blog for the Royal Society:

Safe, secure and robust AI for safety-critical systems | Royal Society

Tags

AI

This article was published on