John Salvatier

I'm a researcher at AI Impacts. I find data to improve our understanding of the likely long run impacts of powerful machine intelligence.

I think humanity's long run future is overwhelmingly important, and mostly ignored. Humanity is not very good at prioritizing the most important problems.

Because of that, I care a lot about rationality and improving humanity's ability to recognize and strategically deal with big important problems, like death, suffering, and existential risk.

My work

More of my writings. More of my projects.

Selected publications

Owain Evans, Andreas Stuhlm├╝ller, John Salvatier, and Daniel Filan. Modeling Agents with Probabilistic Programs. http://agentmodels.org.

Abel D, Salvatier J., Stuhlmüller A., Evans O. (2016) Agent-Agnostic Human-in-the-Loop Reinforcement Learning. Future of Interactive Learning Machines Workshop at NIPS 2016

Kreuger D., Leike J., Evans O., Salvatier J. (2016) Active Reinforcement Learning: Observing Rewards at a Cost. Future of Interactive Learning Machines Workshop at NIPS 2016

Salvatier J., Wiecki TV., Fonnesbeck C. (2016) Probabilistic programming in Python using PyMC3. PeerJ Computer Science 2:e55

More publications

jsalvatier@gmail.com
github.com/jsalvatier