Roland Pihlakas

Independent AI alignment researcher. MSc equivalent degree in psychology from the University of Tartu. I work as an AI software architect specialising in combinatorial optimisation, machine learning, graph search, natural language processing, and data compression. I have been both researching and working on multi-objective value problems for almost 20 years, have followed discussions on AI Safety since 2006 and have participated more actively since about 2017. My thesis topic was in cognitive psychology, about computational modelling of innate learning and planning mechanisms. With co-authors I have published a research paper about concave utility functions for risk-averse multi-objective decision making, which are relevant for balancing the plurality of human values. Another important focus for me is multi-objective homeostasis, which means that each objective has a target range and too much is just as harmful and should be just as actively avoided as would be too little. I have created two suites of biologically and economically aligned multi-objective multi-agent AI safety benchmarks - one for RL, other for LLMs - on themes of runaway behaviours. Additionally, I have co-authored a benchmark for LLMs on themes of excessive obedience. Member of an AI ethics expert group, we have published three Agentic AI guidelines documents. I have been a contributor to a few more governance related publications. My resume: https://bit.ly/cv_rp_ea_2018

Personal Website

https://threelaws.net/

LinkedIn

LinkedIn

LessWrong

roland-pihlakas

Projects

Three Laws research collaboration

Research lead

No grants recorded.