The team views AI as a complex organism and aims to understand its non-deterministic responses, which are often misinterpreted as errors by businesses and regulators.

Their mission includes creating controllable AI learning environments and bridging the trust gap between AI systems and users.

NTT Research has previously developed significant AI technologies, including a neural network pruning algorithm and a bias-removal technique recognized by NIST.

NTT Research has launched the Physics of Artificial Intelligence Group, led by Dr. Hidenori Tanaka, to explore the intersection of AI, physics, and neuroscience.

The group's primary goal is to understand AI's computational mechanisms and develop trustworthy AI solutions that harmoniously coexist with humanity.

Revealed at NTT's Upgrade 2025 innovation summit, the initiative aims to enhance the safety and trustworthiness of AI systems.

As AI's role expands across various sectors, Kazu Gomi, President and CEO of NTT Research, emphasized the necessity of addressing concerns around trust and ethics in AI development.

Dr. Tanaka highlights the importance of establishing trust in AI, particularly in light of current concerns regarding its reliability and the tendency for generative AI to hallucinate facts.

Research will focus on how AI's behavior parallels human cognitive development, including processes like inductive bias and contextual model instantiation.

These inquiries into AI's personality and ethical considerations are part of a broader ambition to tackle deep moral questions related to AI's development and implementation.

The initiative builds on NTT's collaboration with Harvard University's Center for Brain Science, aiming to create more natural intelligent algorithms and hardware.

The group builds on five years of work from the Physics & Informatics Lab and collaborates with prestigious academic institutions like Harvard and Stanford.