Marc-Everin Carauleanu

Marc-Everin Carauleanu is an AI safety researcher and Chief Scientist at AE Studio, based in Oxford, United Kingdom. He holds a BSc in Artificial Intelligence from Oxford Brookes University. He has previously been a Summer Research Fellow at the Stanford Existential Risk Initiative (SERI) in 2021, a Student Researcher at the Center for AI Safety (CAIS) in 2022, and received an LTFF grant in 2021 to write a paper on cognitive and evolutionary insights for AI alignment. His primary research agenda focuses on operationalizing "self-other overlap" — a concept from the cognitive neuroscience of empathy — as a mechanism to reduce deceptive behavior in AI systems. He co-authored the paper "Towards Safe and Honest AI Agents with Neural Self-Other Overlap" (arXiv:2412.16325), which was presented at the NeurIPS 2024 Safe Generative AI Workshop and demonstrated significant reductions in deceptive responses across multiple large language models.

No endorsements yet.

Personal Website

-

LinkedIn

LinkedIn

Twitter / X

-

LessWrong

marc-everin-carauleanu-carauleanu

EA Forum

-