When was the last time your smartphone outsmarted you? Did it scan a barcode on a billboard, remind you to do your laundry or recognize a new song on the radio? Recent advances in technology make it seem like our computers and phones could quickly surpass our brains’ computing power, if they haven’t already. But Kwabena Boahen, associate professor of bioengineering, would quickly argue that there are faults in that belief.
Boahen is the leader of the Neurogrid project in neuromorphic engineering at Stanford, where researchers are working to create a computer system that works more like the human brain — efficiently and with energy consciousness.
“There is a certain elegance that I see in the brain that I want to be able to capture,” Boahen said.
A human brain generates information traffic 1,000 times greater than the global Internet, yet it is able to operate on 10 watts of power. Current computers would require about 10 megawatts of energy to compute like our brains do, according to Boahen.
“Right now, when we model the brain with a computer it is very inefficient, and it takes a supercomputer to carry out the computations,” Boahen said. “With Neurogrid, we are not trying to build a new kind of supercomputer. We are creating a machine that solves just one kind of problem.”
That problem is how to efficiently model and study the human brain.
Neurogrid is a collection of silicon chips with transistors that directly simulate the electrical current flowing across a neuron’s membrane in the brain. The transistors are connected, modeling neuron interactions and networks. The hardware is coupled with three layers of software, allowing researchers to model different cells in the brain and their electrical activity. Just 16 small chips assembled on a board can simulate 1 million neurons.
Boahen has been working at Stanford since 2006, but his interest in the intersection of computing and neuroscience came from influences far earlier in his life.
“I was one of those guys who was always hacking stuff as a kid,” Boahen said. “I got my first computer when I was a teenager, more like a glorified calculator really. But I didn’t want to take it apart, it was intimidating.”
He turned to the library to figure out how computers work and came to realize that they were actually quite inefficient.
“I thought it was so ridiculous: you start with ones and zeros and you have to constantly go fetch things from the memory,” Boahen said. “So I thought there had to be a better way than all of that, but at the time I didn’t really know much about how the brain worked.”
When Boahen was an undergraduate in electrical engineering at Johns Hopkins University, the scientific world was interested in neural networks, or how neurons in the brain interact and communicate signals. He found this area of neuroscience intriguing, and this interest stayed with him through his studies at the California Institute of Technology and work at the University of Pennsylvania to his position at Stanford. He considers the development and implementation of neuromorphic technologies to be his true professional calling and plans to dedicate his career to the field.
Boahen and his team recently received a Transformative Research Project award from the National Institutes of Health to research potential applications for Neurogrid technologies. He believes that neuromorphic engineering could be useful in the healthcare field, namely for improving the use of artificial limbs. Boahen cited the many survivors of foreign wars who return home with serious injuries or lost limbs as potential beneficiaries of this technology.
With this grant, Boahen and his team will work during the next five years to develop a chip that receives signals from the human brain and transmits them to wirelessly control a robotic prosthetic limb. The computation required for these tasks is complex, and for such a chip to be implanted into a patient’s brain, it must have extensive computing capabilities in a small, energy-efficient package. This is where Neurogrid technology comes into play — Boahen’s theoretical developments in neuromorphic devices would allow patients to live more independently, giving them high-powered technology in the most efficient form possible.
Boahen’s research is inspirational to many at Stanford, and the potential applications of his work touch many disciplines.
“[The lab’s work is] paving the way for novel computation models that might help us overcome the memory bandwidth limitation faced today in general purpose computing,” said second-year electrical engineering master’s student Swadesh Choudhary, who works in Boahen’s lab.
“He strives for perfection in everything he does and motivates his students to do the same,” Choudhary added.
While his list of accomplishments is long even early in his career and the implications of his research are potentially life changing for many, Boahen remains grounded in the spirit of pure scientific inquiry.
“Me, I’m just curious, I like the intellectual challenge,” he said. “Figuring out how stuff works…that’s satisfying.”