NTT Analysis Launches New “Physics of Synthetic Intelligence” Group to Tackle AI’s Black Field Problem

NTT Research has introduced the launch of its new Physics of Synthetic Intelligence Group, an unbiased initiative spun out from its Physics & Informatics (PHI) Lab. The group was launched as a part of a broader unveiling at NTT’s Upgrade 2025 innovation summit, reflecting rising international curiosity in constructing safer, extra reliable AI methods. The initiative goals to grasp the interior workings of synthetic intelligence, very similar to classical physics as soon as decoded the conduct of falling apples.

NTT Analysis President and CEO Kazu Gomi, keynote speaker at Improve 2025

“We all know that AI can let you know necessary issues… however we don’t know precisely how that occurs. So the notion of Physics of AI for us is: let’s uncover and perceive how AI works.” – mentioned Kazu Gomi, CEO of NTT Analysis

Gomi acknowledged AI’s super potential to rework industries and each day life but additionally addressed key considerations that might hinder its progress, significantly power consumption, privateness, and trustworthiness. With a staff of over 50 world-class researchers and a mission to “improve actuality,” NTT Analysis is doubling down on daring concepts that will take years to commercialize however promise to reshape the technological panorama in profound methods“. As Gomi famous, “We need to prepare for the long run… and thru science, we will make an enormous factor out of it.”

Led by NTT Analysis Scientist Dr. Hidenori Tanaka, an knowledgeable in physics, neuroscience, and machine studying, the brand new Physics of Synthetic Intelligence group seeks to demystify the “black field” nature of synthetic intelligence—significantly how AI fashions study, course of data, and make choices. The group builds on 5 years of foundational work by the PHI Lab and continues a detailed collaboration with tutorial establishments, together with Harvard College’s Middle for Mind Science, Princeton College, and Stanford College.

The brand new group’s mission is threefold: to deepen understanding of AI’s inside mechanisms; to create experimentally controllable studying environments for AI, impressed by physics methodologies; and to bridge the belief hole between AI methods and human customers. The goal isn’t just technical perception but additionally moral AI integration, achieved by transparency relatively than reactive fine-tuning.

NTT Analysis’s earlier work features a extensively cited neural community pruning algorithm, a bias-removal method for big language fashions acknowledged by NIST, and new research into how AI kinds conceptual understanding. The group’s analysis may affect future designs of extra pure, environment friendly AI methods by leveraging similarities between organic intelligence and synthetic neural networks.

As AI’s affect continues to broaden throughout sectors, from healthcare to infrastructure, NTT Analysis President and CEO Kazu Gomi emphasised the significance of belief and ethics in AI improvement. “The brand new group goals to demystify considerations and bias round AI options to create a harmonious path ahead for the coexistence of AI and humanity,” he mentioned.

Dr. Hidenori Tanaka, Scientist, NTT Analysis, PAI Group, NTT Analysis

Dr. Hidenori Tanaka offered his imaginative and prescient on the rising subject he calls the “Physics of AI,” emphasizing that synthetic intelligence will inevitably surpass human intelligence in lots of dimensions. Somewhat than resisting this transformation, he urged a give attention to creating concord between AI and humanity by deep understanding. Drawing parallels to previous scientific revolutions like thermodynamics, Tanaka argued that AI represents a brand new scientific frontier—one the place we will instantly observe and mathematically examine intelligence itself.

He contrasted dystopian Western narratives of AI with Japan’s extra optimistic cultural views, highlighting the potential for AI to be built-in as a collaborative societal entity. Tanaka harassed the significance of redefining conventional scientific domains—like physics, neuroscience, and psychology—to accommodate the complexities of AI methods. He defined that AI fashions not simply as instruments however as reflections of human cognition and personalities, with actual affect on societal constructions and particular person worldviews.

Tanaka advocated for accountable engineering of AI’s social function, cautioning towards its unchecked reinforcement of polarization. Finally, for him, AI is just not as a black field, however a dream field—a clear, manipulable system that may assist us higher perceive ourselves and construct wiser, kinder societies. “AI is just not a black field—it’s a dream field for the science of intelligence.” mentioned Tanaka.

Filed in Announcements >Computers >Robots. Learn extra about AI (Artificial Intelligence) and Ntt.