Correlation does not imply causation. Using AI to discover new materials is becoming more common, but data-driven approaches sometimes rely heavily on correlations and make testing AI-based predictions difficult. A new perspective paper published by Tohoku University researchers developed a framework for incorporating physics-based principles into AI to generate more interpretable, testable, and meaningful predictions.

Inspired by the octopus’s neural connections, electrical contacts enable a continuous artificial muscle to bend at multiple, precisely controlled points, paving the way for compact soft robots capable of navigating, inspecting, and operating in confined spaces.

On social media, Professor Eiryo Kawakami humorously introduces himself as a ‘Triple Dr.’ —a medical doctor (Dr.), a PhD holder (Dr.), and a drummer (Dr.). Behind this playful self-introduction lies a serious mission: advancing Artificial Intelligence (AI) and data science technologies for the early detection and prediction of disease. Professor Kawakami serves as the Director of the Center for Artificial Intelligence Research in Therapeutics, while also playing a central role in the management meetings of the Chiba University Data Science Core (DSC). Balancing these responsibilities with his research and personal interests makes for a busy but stimulating life.

“I enjoy intellectually exploring in many different directions,” he says. We spoke with him about his vision for the future of medicine powered by AI.

When learning a new motor skill, brain connections are reshaped to refine the timing of movements. However, scientists at MPFI have shown that rather than a brain-wide overhaul, synaptic changes in just a single class of neuron are needed to precisely adjust the timing of motor behavior. This finding suggests that with further research, scientists may be able to pinpoint specific learning deficits to precise neuronal cell types.