IFFS PLAY

Research seminar with Anders Sandberg, Ph.D. in computational neuroscience, researcher at IFFS and Senior Research Fellow at the Future of Humanity Institute, Oxford University. Abstract A key problem for the world is global coordination failures with disastrous consequences, such as war, famines, pandemics, etc. There have long been proposals for governance systems to deal with these failures, from Hobbes’ Leviathan to enlightened world government to Bostrom's Singleton. The key idea is to have a unified entity that can both resolve a coordination issue, and have the power to enforce such a solution. We argue that law plays the role of a "natural intelligence singleton", a distributed system that ensures regularity of action in society, making it instrumentally rational to coordinate and ensuring (some) prosperity and stability. This is not just metaphorical: we claim each human society exists within an artificial system of collective extended cognition that solves coordination problems that individually are too hard to solve. However, development of AI may lead to distributed non-human and human-independent systems of cognition that can emergently disempower humans (alignment failure) or make it rational for humans to hand over their autonomy to such systems ("the serfdom conclusion"). We discuss whether there are ways of combining the coordination and prosperity benefit of AI singletons with human flourishing and autonomy. Do you wish to get reminders about our research seminars? Subscribe here: https://www.iffs.se/en/about-us/newsletter/