Brain-like Computing in an Organic Molecular Layer
–Toward Intelligent and Creative Computers–
National Institute for Materials Science
Michigan Technological University
National Institute of Information and Communications Technology
A group led by Dr. Anirban Bandyopadhyay of NIMS, in joint work with Michigan Technological University and NICT, realized the world's first "evolutionary circuit" with processes similar to those of the human brain.
Abstract
- A group led by Dr. Anirban Bandyopadhyay of the Advanced Scanning Probe Microscopy Group, Advanced Nano Characterization Center (Managing Director: Daisuke Fujita) of the National Institute for Materials Science (President: Sukekatsu Ushioda), in joint work with Michigan Technological University (United States) and the National Institute of Information and Communications Technology (President: Hideo Miyahara), realized the world’s first "evolutionary circuit" with processes similar to those of the human brain.
- In the research announced here, the group created a process in which information-processing circuits self-evolve in a manner similar to human neurons, and thus can solve more complex problems than is possible with conventional computers.
- The molecular processor used in this circuit has the following distinctive features:
- Massively parallel processing. Whereas the world’s fastest existing supercomputers process bits sequentially in their respective channels, the newly-developed circuit is capable of instantaneously processing up to 300 bits at one time in parallel.
- Self-healing if a defect occurs. The self-organizing ability of the organic molecular layer provides a self-healing property not found in any existing computer. Furthermore, if a neuron is lost, its functions are taken over by another circuit.
- Intelligent monolayer. The fact that this molecular layer possesses intelligence has been confirmed, realizing the prediction of Alberto Cerdi in "Monolayers with an IQ" (2008).
- In order to demonstrate these unique features, the group simulated two natural phenomena, diffusion of heat and the evolution of cancer cells.
It is considered possible to solve problems for which computer algorithms are unknown, such as prediction of natural disasters and the outbreak of diseases.
- This research will be announced in advance in the online edition of Nature Physics, which is a sister publication of Nature, at 2:00 April 26 (Monday), Japan time (18:00 April 25, local time in London).
