NTT Research has launched the Physics of Artificial Intelligence Group, a new basic research unit spun out from its Physics & Informatics (PHI) Lab. The group builds on a five-year foundation of interdisciplinary research at the intersection of physics, neuroscience, and machine learning, with a mission to enhance understanding of the internal workings of AI systems. Led by NTT Research Scientist Dr. Hidenori Tanaka, the group will focus on developing insights into AI’s “black box” operations to improve trust, safety, and ethical design in future AI technologies.
The new group aims to explore how artificial intelligence systems accumulate knowledge, make decisions, and mirror biological intelligence. By applying techniques from physics, the team will construct controlled environments to study AI behavior and develop systematic approaches for bias reduction, concept learning, and secure decision-making. Notable past contributions from the group include a neural network pruning algorithm with over 750 citations and a bias-removal method for large language models recognized by the U.S. National Institute of Standards and Technology (NIST).
Looking ahead, the group will deepen its collaborations with academic partners including Harvard, Princeton, and Stanford to advance research in AI safety and energy-efficient computation. It will also explore synergies with emerging technologies from the PHI Lab, such as photonics-based computing and thin-film lithium niobate (TFLN) devices, to build more human-aligned, power-conscious AI systems.
• NTT Research launches Physics of Artificial Intelligence Group to study and demystify AI decision-making
• Group led by Dr. Hidenori Tanaka, building on work from the PHI Lab and academic collaborations
• Focus areas include bias mitigation, explainability, biological parallels, and energy-efficient computing
• Previous contributions include a widely cited pruning algorithm and AI bias-reduction methods
• Partnerships include Harvard Center for Brain Science, Stanford, and Princeton
• Group will contribute to ethical, interpretable, and safe AI through interdisciplinary research
“With the emergence of this group, we have a path forward to understanding the computational mechanisms of the brain and how it relates to deep learning models,” said Hidenori Tanaka, Head of the Physics of Artificial Intelligence Group at NTT Research.
