A joint research team from NIMS and Tokyo University of Science has successfully developed an advanced artificial intelligence (AI) device that performs brain-like information processing using supramolecular reservoir computing. This innovation uses the molecular vibrations of a certain number of organic molecules. By applying this device to predict blood sugar levels in diabetic patients, it significantly outperformed existing AI devices in terms of prediction accuracy.

With the proliferation of machine learning programs in various industries, the demand for AI devices that not only have high computing power but also low power consumption and miniaturization is increasing. Research has moved towards computing physical reservoirs using physical events represented by neural information processing materials and devices. One remaining challenge is the relatively large size of available materials and devices.

Groundbreaking development in reservoir computing

The research pioneered the world’s first application of physical reservoir computing, which works on the principle of surface Raman scattering using the molecular vibrations of only a few organic molecules. Information is presented using ion gating, which modulates the adsorption of hydrogen ions on organic molecules (p-mercaptobenzoic acid, pMBA) using voltage.

Changes in the molecular vibrations of pMBA molecules due to the adsorption of hydrogen ions realize the memory function and nonlinear transformation of the signal shape for calculation. Using a rare population of pMBA molecules, it examined changes in a diabetic patient’s blood sugar level over approximately 20 hours and was able to predict subsequent fluctuations over the next 5 minutes with an error reduction of approximately 50% compared to the peak value. Accuracy achieved with similar devices to date.

The results of this study show that a minimal number of organic molecules can perform computer-comparable calculations efficiently. This technological breakthrough in complex information processing in minimal materials and small spaces provides significant practical advantages. This paves the way for low-power AI terminal devices that can be integrated with a variety of sensors, paving the way for widespread industrial use opportunities.