News

Hidenori joined the Cambridge Boston Alignment Initiative (CBAI) Summer Research Fellowship in AI Safety as a mentor, advising fellows on research at the interface of collective intelligence, multi-agent systems, and AI safety.

NTT Research announced that Harvard’s Dr. Mai Uchida has joined the Physics of AI Group. Her work connects child psychiatry, neuroscience, and AI, strengthening the group’s research on mental health, education, and safe human-AI interaction.

Read the full announcement at NTT Research.

Five papers from the group were accepted to ICLR 2025, including “In-Context Learning of Representations,” “Competition Dynamics Shape Algorithmic Phases of In-Context Learning,” “Dynamics of Concept Learning and Compositional Generalization,” and a percolation model of emergence in transformers trained on a formal language.

See the Publications page for full citations and links.

Read the full announcement at NTT Research.

Hidenori joined the MATS (ML Alignment & Theory Scholars) program as a mentor, advising scholars on projects at the interface of mechanistic interpretability, learning dynamics, and alignment theory.

Our paper Emergence of Hidden Capabilities: Exploring Learning Dynamics in Concept Space (Park et al., 2024) was selected as a Spotlight at NeurIPS 2024.

Read the full announcement at NTT Research.

We launched the CBS–NTT Program in Physics of Intelligence at Harvard’s Center for Brain Science — a long-term collaboration with NTT Research to build a community of physicists, neuroscientists, and AI researchers working on a science of intelligence.

Read the full announcement in the Harvard Gazette.