Meeting KI-Postdocs: ‘Taming Superintelligence’ "Governing global change through AI safety" (Dr. Apolline Taillandier) This presentation focuses on the recent history of “AI safety,” understood as a set of strategies for mitigating long-term AI risk. AI safety brings together computer science and policy expertise: while some AI safety researchers have redesigned algorithmic training methods for ensuring control over the goals or behavior of “human-level” AI, or systems with “artificial general intelligence,” (AGI) AI safety policy experts have developed scenarios and metrics of AI progress in order to anticipate and prepare for uncertain technological breakthroughs through specific models of governance. I show how AI safety expertise intersects transhumanist debates about the singularity and the possibility to predict and tame the emergence of a posthuman kind of intelligence; economic models of rational behavior in computer science; and a realist model of global security >as a product of interest-based cooperation and alignment. https://www.uni-bonn.de/en/research-and-teaching/research-profile/transdisciplinary-research-areas/tra-4-individuals/events-1/copy2_of_what-machines-shouldnt-do-part-of-the-ai-research-group https://www.uni-bonn.de/++resource++plone-logo.svg

"Governing global change through AI safety"
(Dr. Apolline Taillandier)
This presentation focuses on the recent history of “AI safety,” understood as a set of strategies for mitigating long-term AI risk. AI safety brings together computer science and policy expertise: while some AI safety researchers have redesigned algorithmic training methods for ensuring control over the goals or behavior of “human-level” AI, or systems with “artificial general intelligence,” (AGI) AI safety policy experts have developed scenarios and metrics of AI progress in order to anticipate and prepare for uncertain technological breakthroughs through specific models of governance. I show how AI safety expertise intersects transhumanist debates about the singularity and the possibility to predict and tame the emergence of a posthuman kind of intelligence; economic models of rational behavior in computer science; and a realist model of global security >as a product of interest-based cooperation and alignment.