Revolutionary AI Device Mimics Human Brain With Few-Molecule Computing
In a groundbreaking advancement, scientists have made a significant leap forward in artificial intelligence (AI) technology by developing a new AI device capable of executing brain-like information processing with just a few molecules. This device, established through the collaboration between the National Institute for Materials Science (NIMS) and Tokyo University of Science, uses the molecular vibrations of a handful of organic molecules to perform computations, marking a substantial breakthrough in the realm of compact AI devices.
This innovative AI device employs a concept known as few-molecule reservoir computing. The promising field of physical reservoir computing takes advantage of natural physical phenomena for neural information processing, offering a potential solution to the demand for devices that are not only powerful but also efficient and compact. The newly developed device utilizes surface-enhanced Raman scattering to capture the molecular vibrations of a sparse number of organic molecules—specifically, p-mercaptobenzoic acid (pMBA)—to process information.
The operation of this cutting-edge device is both simple and ingenious. By applying a voltage, hydrogen ions’ adsorption onto the pMBA molecules is modulated, a process known as ion-gating. The subsequent changes in the vibrations of these molecules, induced by the adsorption of hydrogen ions, act as a means for memory and nonlinear waveform transformation crucial for computation. Through this method, the device has successfully learned and predicted blood glucose level changes in diabetes patients with unprecedented accuracy, reducing the error rate by about 50% compared to currently available AI devices.
The implications of this research are far-reaching. By demonstrating that a minimal assembly of organic molecules can perform complex computations typically requiring a computer, the scientists have opened new avenues for creating AI devices that are not only highly efficient but also significantly smaller and more power-effective than existing models. This can potentially lead to the development of low-power AI terminal devices tailored for various industrial applications, significantly benefitting sectors that require compact, yet powerful AI solutions.
The study, entitled “Few- and single-molecule reservoir computing experimentally demonstrated with surface-enhanced Raman scattering and ion gating,” was published in the journal Science Advances and marks a collaborative effort led by Daiki Nishioka, a JSPS Research Fellow at Tokyo University of Science, along with Takashi Tsuchiya and Kazuya Terabe of the Ionic Devices Group at NIMS. Their work falls under the project “Nano Materials for New Principle Devices,” focusing on the “Creation of Ultrafast Iontronics,” sponsored by JST PRESTO.
This novel approach to AI marks a pivotal moment in the evolution of artificial intelligence technologies. The ability to perform complex computational tasks with a minimal number of molecules not only challenges our traditional perceptions of computing hardware but also lays the groundwork for future innovations in AI devices. As we continue to explore the limits of molecular computing, the boundary between biological and artificial intelligence becomes ever more blurred, heralding a new era of technological advancements.
Reference: Nishioka, D., Shingaya, Y., Tsuchiya, T., Higuchi, T., & Terabe, K. (2024). “Few- and single-molecule reservoir computing experimentally demonstrated with surface-enhanced Raman scattering and ion gating.” Science Advances, 28 February 2024. DOI: 10.1126/sciadv.adk6438
