About

I am Mateusz Bagiński. I go by MatthewBaggins in most online places. This is my very WIP website. More stuff will be added to it over the next few weeks/months/years. For now, here is my Linktree.

My purpose is to push the world as much as possible in the direction of the good future, primarily by doing stuff that has a good chance of:

• averting AI X-risks
• ensuring positive effects of AI
• improving rationality/[coordination ability]/sanity of Humanity/Civilization

My current primary focus is an agent-foundations-theoretic investigation of the core drivers of goal-directed mental activity. The purpose is to find some foundational threads that will lead us to an understanding sufficient to «“align AGI”» or at least put reasonable bounds on what may come out of weakly superhuman AI systems.

Other interests:

• High actuation spaces
• Formal methods for alignment (e.g. Davidad’s Open Agency Architecture/Safeguarded AI)
• Developmental interpretability/Singular Learning Theory
• Naturalized induction (e.g. Logical Induction, Infra-Bayesianism)
• Philosophy of probability
• Category theory
• Evolutionary biology (and [evolution/natural selection] more broadly than biology)
• Complex systems

If you want to talk to me, send me an email. I’m in Belgrade until mid-June, after which I will probably spend 3 months in London. What happens next is as-of-today-TBD.