The tech giant IBM research center in Zurich created 500 of them to simulate a signal transfer similar to how the process works in an organic brain.
IBM researcher builds artificial neuron similar to how the process works in an organic brain
As other research in artificial signaling demonstrate, the real milestones are had when elements can be shrunk down to microscopic scale and still work. That's what makes IBM's accomplishment significant: their faux neurons are built out of well-known materials that can scale down to a few nanometers but can still activate with low energy, points out Ars Technica.
Organic neurons have membranes acting as signal gates that take a certain amount of energy to absorb. In the IBM version, that role is taken by a square of Germanium-Antimony-Tellerium (GST), a common ingredient in optical disks. Heat the GST enough and it changes its physical phase, from an amorphous insulator to a crystalline conductor. In other words, signal passes through when the faux membrane is hit with enough electricity to change into its crystal phase, then it resets to its amorphous one.
We have been researching phase-change materials for memory applications for over a decade, and our progress in the past 24 months has been remarkable, said IBM Fellow Evangelos Eleftheriou. In this period, we have discovered and published new memory techniques, including projected memory, stored 3 bits per cell in phase-change memory for the first time, and now are demonstrating the powerful capabilities of phase-change-based artificial neurons, which can perform various computational primitives such as data-correlation detection and unsupervised learning at high speeds using very little energy.
But the scientists needed the artificial neuron to have another characteristic of its organic counterpart: stochiasticity, or some randomness in when signals will fire. IBM says its neurons achieve this because its GST membranes never reset to the same configuration. This lets groups of them unexpectedly accomplish things that they could not if their results were perfectly predictable.
With these neurons, scientists may be able to create computers mimicking the efficient, parallel processing design of organic brains and apply its style of approach to decision-making and processing sensory information, suggests Ars Technica. But as they point out, constructing it might be the easy point: writing software for that kind of setup will be another challenge entirely.