A Two-Way Street
In the three years we’ve been publishing this newsletter, we’ve reported on a number of ways in which electronics technology can help the nervous system. In our article on neural-silicon hybrids this month [see p1], we’re pleased to report the nervous system is ready to return the favor. Biologically inspired processors—perhaps even neurons themselves—may soon play a role in the next generation of computational devices.
Much of the progress in neural-silicon hybrid technology has taken place at the University of Southern California. Ted Berger, who directs the Center for Neural Engineering at the university, assembled an interdisciplinary team of neuroscientists, biomedical engineers, electronic circuit designers, molecular biologists, and computer scientists to build a cortical prosthesis to replace hippocampal neural cells lost to disease or injury. Berger’s work was a major reason why USC was selected by the National Science Foundation to house the engineering research center devoted to biomimetic microelectronics [NBR Jan04 p1].
Berger’s initial goal, and one of the three primary thrusts of the new biomimetic center, is to treat cognitive impairments such as memory loss with an implantable neural prosthesis. But in the process of constructing this device, Berger’s team has made tremendous strides in neurocomputational technology. The current hardware, though likely years away from human implantation, has already demonstrated many of the nonlinear dynamic and adaptive properties of real neurons and can perform basic pattern recognition functions. The USC researchers have also made impressive progress in developing microelectrode arrays that are designed to the specific cytoarchitechure of the brain region in which they will be implanted. This “neuromorphic” approach should enable the implant to blend in with neural tissue in a functionally seamless and biocompatible manner.
Ultimately, we expect many more benefits from this biomimetic engineering effort than a hippocampal prosthesis. The process of studying the input-output relationships of electrical signals interacting with a neuron over space and time has given the USC team an enormous head start in building computational devices that incorporate true neural network concepts. Once commercialized, biomimetic or neural-slicon hybrid processors will be able to accomplish far more than the first generation of neural network systems that emerged in the last two decades based on von Neumann computer architectures. Perhaps it’s not too much of a stretch to envision that neural-silicon hybrid processors will form the basis for the next generation of massively parallel computer architectures.
The interface between neuron and electronic device is truly a two-way street that not only offers a bidirectional channel of communication between man and machine, but also a glimpse at how both will process information in the future.
Editor and Publisher