Thursday, May 02 - Toulouse - Jean François Bonnefon and a wide team of academic and industrial researchers have published a new framework article in Nature on AI behaviour. They demonstrate that studying machine behaviour necessitates a broad scientific research agenda, well beyond the boundaries of computer science.
Machines powered by Artificial Intelligence mediate our social, cultural, economic and political interactions. While machines can benefit society there is also a fear of potential loss of human oversight over the latter as well as potentially harmful misuse.
In this new publication, over 20 interdisciplinary scientists, roboticists, engineers and academic researchers from Canada, France, Germany, the UK and the USA as well as industrials researchers from Microsoft, Facebook and Google came together to set out an unprecedented framework on how to best study machine behavior. Their work explored the technical, legal and institutional constraints, leading to the creation of a framework which shows that understanding and monitoring machine behaviour is necessarily a collaborative and interdisciplinary endeavor, in order to fully grasp the stakes and benefit from the insights of all fields.