Diving into the world of artificial intelligence (A.I.) can often resemble stepping into a complex “black box” where the mechanisms at play are not immediately clear. Yet, a dedicated group of researchers from Harvard is striving to shed light on this technology and reveal the intricacies of A.I. learning.
During the recent Upgrade 2025 conference held in San Francisco, NTT Group, a prominent Japanese IT services firm, introduced the Physics of Artificial Intelligence (PAI) Group at Harvard University’s Center for Brain Science. The aim of this group is to uncover the essential principles that govern A.I.’s learning and reasoning processes, akin to how the laws of physics clarify the principles of motion and energy.
“To genuinely grasp and address A.I.’s black-box dilemma, it’s essential to explore its psychological and structural dimensions—understanding how it perceives information, makes decisions, and the rationale behind those processes,” stated Hidenori Tanaka, the head of PAI. Tanaka, who is a leading A.I. researcher at Harvard, highlighted the need to move beyond conventional benchmarking approaches to assess A.I.’s abilities.
The PAI team is developing “model experimental systems” to track the way A.I. models learn and reason over time. By assembling a variety of datasets spanning subjects such as physics, chemistry, biology, mathematics, and language, they aim to reveal how A.I. interprets novel concepts and maintains knowledge. This methodology will not only improve A.I. systems but also help reduce biases and inaccuracies in future technologies.
With a Ph.D. in theoretical physics from Harvard, Tanaka is particularly focused on aligning the computational principles of the brain with the laws of physics. PAI, which evolved from NTT Research’s Physics and Informatics Lab, collaborates with distinguished researchers from institutions like Harvard, Princeton, and Stanford to push the boundaries of A.I. research.
In an era where A.I. increasingly influences our lives, comprehending its learning mechanisms is crucial for unlocking its complete potential. The expedition to decode the enigma of artificial intelligence is just getting underway, and with every breakthrough, we edge closer to a future where technology and humanity can coexist in harmony.
