Can we build an elegant mathematical theory of reasoning and planning as a blueprint for artificial (super)intelligence? What does it even mean for a theory to be elegant? What implications does this have for our own thinking and the scientific process?
Universal Algorithmic Intelligence (UAI) is a rigorous mathematical approach to answering these questions using computability theory and probability. Starting from a model of idealized reasoning without computational constraints, UAI researchers work “backward” toward practical algorithms and implementations. This means that there is a lot of philosophical work to be done establishing (and/or contending) both the optimality of UAI notions (e.g. are learning rules such as Solomonoff induction the best possible?) and their applicability to real minds (e.g. can biological and/or artificial intelligences usefully approximate Solomonoff induction in practice?). There is also engineering work to be done creating UAI-inspired AI models as proofs of concept!
See the “About” page for details about UAI and related topics of interest. You can also ask questions here.
Subscribe below to be notified of community blog posts including research discussion and event announcements such as new conferences and reading groups.
We hold regular meetings online. Please reach out to an organizer (e.g. colewyeth@gmail.com or see “Contacts”) for a meeting link. The regular meetings are for research and paper discussions while the reading group “Introduction to Universal Artificial Intelligence” is a good on-ramp for new students and interested researchers from other fields.
Universal Algorithmic Intelligence 