Paul Riechers
Co-founder, SimplexAI Safety; Research Lead, Astera Institute
-
Paul Riechers co-founded and leads research at Simplex, an AI interpretability organization discovering the representations, abstractions, and mechanisms of neural networks. By building the fundamental science of intelligence, Simplex aims to elevate the conversation around AGI/ASI, and to increase optionality for better futures. Paul earned his PhD in theoretical physics and MS in electrical and computer engineering at UC Davis, and was then a research fellow for five years at NTU in Singapore. He co-founded the Beyond Institute for Theoretical Science (BITS), served as a Senior Fellow at IPAM’s Mathematics of Intelligences program at UCLA, and has mentored technical AI safety work at MATS, PIBBSS, and ILIAD.
Previously, Charles worked at Posit, PBC, where he created the Quarto scientific publishing system. Charles also founded Lose It! co-founded Onfolio, and was a VP of Engineering at Allaire Corp. -
Build upon Simplex's mathematical framework to anticipate and discover how neural networks build and use geometric arrangements of context embeddings. Discover how AI systems carve up their world into parts and abstractions. Find how these compose, and the implications for behavior and possible interventions.
Perform theory-informed mechanistic interpretability on either (i) small transformers trained on curated synthetic data, and/or (ii) LLMs or multimodal models.
-
Demonstrated track record of excellence in research and/or ML.
Cares deeply about interpretability, AI safety, and understanding intelligence.
Comfortable with mathematical modeling, and discussing mathematical frameworks.
Great experimental design, implementation, and analysis of experiments.
Clear and concise communicator.