Constructive AI Alignment
An agenda for the coevolution of AI systems and people.
A personal research agenda between me and my partner, Max Kanwal, which started in our undergraduate years at UC Berkeley after attending workshops and courses on AI safety. We are proposing a more interdisciplinary, systems-level approach to AI alignment which accounts for the interactive and dynamic nature of interaction between humans and machines. We’ve spelled out our aims and approach with our paper Constructive Alignment and are working towards more technical research agendas.