NAI Senior Members are active faculty, scientists and administrators "who have demonstrated remarkable innovation-producing technologies that have brought or aspire to bring, real impact on the welfare of society", according to the Academy. They have also proved successful in patenting, licensing and commercializing their inventions.
Bipin Rajendran, an associate professor of electrical and computer engineering and expert in nanoscale electronic devices and system design, holds 59 issued U.S. patents. He joins a class of 54 newly elected members from institutions including Yale University, the Texas Heart Institute and the Naval Information Warfare Center.
"Advances in computing have enabled systems with astonishing capabilities that augment and even surpass human capacity in many facets of life, but there is a crucial gap in all of these awe-inspiring artificial systems: the enormous amount of energy they consume to perform their tasks", he noted of technologies such as two-legged humanoid robots developed for search-and-rescue operations in hazardous environments and big data-analytics engines that work alongside doctors to diagnose diseases and suggest treatment plans.
He added: "While the Watson supercomputer from IBM required 85.000 watts to challenge and ultimately vanquish two "Jeopardy!" champions, Watson's conqueror, former U.S. Congressman Rush Holt, relied on a far more efficient machine - the human brain - which functions on a mere 20 watts."
At the heart of these brain-inspired systems are artificial neural networks, which are mathematical models of the networks of neurons and synapses in the brain. In order to endow them with human-like intelligence, vast troves of information are fed to them, and their internal parameters - the strengths of the synaptic connections between the neurons - are adjusted so that they learn the hidden relationships that underlie different parts of the data. A network trained in this fashion can generalize and create meaningful actions when it is presented with similar, but hitherto unseen data. By bridging the efficiency gap, Bipin Rajendran hopes to enable intelligent and agile electronic assistants that can be widely deployed in homes, roads, cities and natural environments.
"While machine-learning algorithms are capable of executing complex tasks such as controlling self-driving cars and interpreting a growing number of languages, use of these algorithms in mobile devices and sensors embedded in the real world requires new technologies that consume substantially less energy", Bipin Rajendran noted.
To date, however, the implementation of brain-inspired algorithms on conventional computers is highly inefficient, consuming huge amounts of power and time. Today's computers are built on an architectural scheme that was developed by John von Neumann in the early 1940s. In these machines, the data storage unit (memory) and the data processing unit (processor) are physically separated, and data continually shuttle back and forth during computations based on well-defined algorithms or programmes. This is vastly different from the architecture in the brain, where logic units (neurons) and memory units (synapses) are seamlessly integrated in a dense 3D network.
"We believe that we can substantially improve computational efficiency by designing systems that seamlessly integrate storage and data-processing functions", he stated. "These novel systems eliminate, at least partially, the need to shuttle data back and forth, as several key computational steps can be implemented while the data resides in the memory."
Last year, Sagnik Basuray, an NJIT assistant professor of chemical engineering who develops novel sensors, diagnostic devices, and drug delivery systems, was elected to the inaugural class of NAI Senior Members.