NTT Research, a division of NTT, presented the Physics of Artificial Intelligence Group, which was formed from a group within the NTT Research Physics & Informatics (PHI) Lab. The new group will continue to improve the “Physics of Artificial Intelligence,” an interdisciplinary approach to understanding AI pioneered by the team in the previous five years.
Dr. Hidenori Tanaka, a physics, neurology, and machine learning expert who previously led the PHI Lab’s Intelligence Systems Group, will take over as head of the Physics of Artificial Intelligence Group.
Early on, the PHI Lab recognized the need of understanding the “black box” nature of AI and machine learning in order to design revolutionary systems with significantly increased computational energy efficiency. With AI evolving at such a rapid pace, questions of trustworthiness and safety have emerged as crucial to commercial applications and AI governance.
The Physics of Artificial Intelligence Group, working with leading academic researchers, aims to address similarities between biological and artificial intelligences, further unravel the complexities of AI mechanisms, and foster trust, resulting in a more harmonious fusion of human and AI collaboration.
The goal is to have a deeper understanding of how AI works in terms of training, information acquisition, and decision making so that we can create coherent, safe, and trustworthy AI in the future.
This approach is similar to what physicists have done for centuries: people recognized how objects move when forces are applied, but it was physics that revealed the specific intricacies of the relationship, allowing humans to construct the machines we know today.
For example, the invention of the steam engine influenced our understanding of thermodynamics, allowing the production of sophisticated semiconductors. Similarly, the efforts of this team will influence the future of AI technology.
The new group will continue to work with Harvard University’s Center for Brain Science (CBS), chaired by Harvard Professor Venkatesh Murthy, and Princeton University’s Assistant Professor (and former NTT Research Scientist) Gautam Reddy. It also intends to work with Stanford University Associate Professor Surya Ganguli, with whom Tanaka has co-authored multiple publications.
Tanaka, NTT Research Scientist Maya Okawa, and NTT Research Postdoctoral Fellow Ekdeep Singh Lubana make up the group’s core personnel. Previous contributions to date include:
- A widely cited neural network pruning algorithm (over 750 citations in just 4 years)
- A bias-removal algorithm for large language models (LLMs), recognized by the U.S. National Institute of Standards and Technology (NIST) for its scientific and practical insights; and
- New insights into the dynamics of how AI learns concepts
The Physics of Artificial Intelligence Group’s future mission consists of three components.
1) It aims to improve our understanding of AI systems in order to integrate ethics from within, rather than through a patchwork of fine-tuning (i.e. compulsory learning).
2) Drawing on experimental physics, it will continue to create systematically controllable AI spaces while gradually observing AI learning and prediction behaviors.
3) It aims to restore confidence between AI and human operators by improving operations and data control.
The Physics of Artificial Intelligence Group takes an interdisciplinary approach to AI, bringing physics, neurology, and psychology together. This strategy goes beyond standard benchmarks, recognizing the importance of promoting goals like fairness and safety that contribute to long-term AI acceptance.
Reducing Energy Consumption of AI Computing Platforms with TFLN Technology
In terms of energy efficiency, several groups in the PHI Lab are already working to lower the energy consumption of AI computing platforms using optical computing and a groundbreaking thin-film lithium niobate (TFLN) technology.
Furthermore, inspired by the large difference in wattage consumed by LLMs versus the human or animal brain, the new group will investigate ways to utilize commonalities between biological brains and artificial neural networks.
PHI Lab Research With Photonics-Based Technology
Since 2019, the PHI Lab has led research into novel ways of computing systems using photonics-based technology. This initiative investigates TFLN-based devices, whereas the Coherent Ising Machine offers new perspectives on complex optimization problems that were previously difficult to address on traditional computers.
Since 2019, the PHI Lab has led research into novel ways of computing systems using photonics-based technology. This initiative investigates TFLN-based devices, whereas the Coherent Ising Machine offers new perspectives on complex optimization problems that were previously difficult to address on traditional computers.
Key Joint Research Agreements
In addition to a joint research agreement (JRA) with Harvard, the PHI Lab has previously collaborated with the California Institute of Technology (Caltech), Cornell University, Harvard University, Massachusetts Institute of Technology (MIT), Notre Dame University, Stanford University, Swinburne University of Technology, the University of Michigan, and the NASA Ames Research Center. Overall, the PHI Lab has published over 150 papers, including five in Nature, one in Science, and twenty in Nature sister journals.
Leadership Comments
“Today marks a new step towards society’s understanding of AI through the establishment of NTT Research’s Physics of Artificial Intelligence Group,” NTT Research President and CEO Kazu Gomi said. “The emergence and rapid adoption of AI solutions across all areas of everyday life has had a profound impact on our relationship with technology. As AI’s role continues to grow, it is imperative we explore how AI makes people feel and how this can shape the advancement of new solutions. The new group aims to demystify concerns and bias around AI solutions to create a harmonious path forward for the coexistence of AI and humanity.”
“The key for AI to exist harmoniously alongside humanity lies in its trustworthiness and how we approach the design and implementation of AI solutions,” Hidenori Tanaka said. “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. Looking ahead, our research hopes to bring about more natural intelligent algorithms and hardware through our understanding of physics, neuroscience, and machine learning.”
To Know More on Physics of Artificial Intelligence Group: CLICK HERE
