Computing inspired by the brain
doi:10.1038/nindia.2015.102 Published online 17 August 2015
Human brains, just like computers, process information. Yet there are profound differences between the two. “We lose brain cells all the time, yet our memory or processing doesn't disappear all at once,” says Mriganka Sur, Professor of Neuroscience at Massachusetts Institute of Technology (MIT), Boston, USA. What’s more, brains don’t need to be programmed and consume as little power as a dim light bulb.
Can we design better computers using a deeper understanding of how the brain works? That’s the question Kris Gopalakrishnan, co-founder of Indian software giant Infosys, wants to explore with a new endowment.
Gopalakrishnan, an Indian Institute of Technology-Madras (IIT-M) alumnus, is funding three endowment chairs of ₹ 10 crores each in his alma mater to help investigate this question. Apart from Sur, two other leading scientists — Partha Mitra, a neuroscientist and professor of biomathematics at Cold Spring Harbor Laboratory in New York, and Anand Raghunathan, a professor of electrical and computer engineering at Purdue University — are occupants of the chairs in IIT-M’s Centre for Computational Brain Research. The trio were in Chennai in late July 2015 to chart a course of action for the Centre.
For half-a-century, progress in the computing industry has been in accordance with Moore’s law, as computing power roughly doubles every two years with more transistors packed onto tinier chips. Also, increasingly powerful — and power-hungry — computers are needed to process the massive amounts of data being generated. For instance, Mitra’s lab at Cold Spring is engaged in mapping mouse brain circuits, where a single brain generates about a terabyte, or a thousand gigabytes, of data.
But our brains process big data all the time — information that’s streaming in through our sensory organs is transformed into a coherent picture of the world. For example, the brain takes the input from the eyes and, among other things, processes colour and shape separately. This makes it a natural parallel processor too. Researchers at the centre will work to reveal some of these principles behind how the brain works differently from computers – how different kinds of neurons code information, how they integrate input from sensory organs and transform it to decisions and actions.
The aim will be to then implement these principles in computers. This could be at the device level by building hardware that emulates neurons. Transistors in today’s computer chips are digital – they have two states, on or off – unlike neurons which can carry electrical signals that vary continuously. Or, says Raghunathan, “You can take today's circuits and architectures but then bring in brain-inspired ideas to write software that run on today's computers more efficiently, perhaps more robustly.”
That’s the field of neuromorphic computing, where physicists, electrical engineers, computer scientists and biologists are involved. The three scientists will collaborate with existing IIT-M faculty members and their graduate students and will also oversee the recruitment of new faculty. “Neuroscience is going through this big data phase now and that’s an opportunity for Indian scientists to engage,” says Mitra. The computing infrastructure required will be provided by a Centre for Data Science, separately funded.
There are other big-money neuroscience efforts worldwide, such as the BRAIN initiative in the U.S., in which both Sur and Mitra are involved. But the focus at IIT-M will be different. “It is to leverage the results coming out from their efforts, and other efforts, to really focus on this feedback between computing and neuroscience,” says Raghunathan.
That feedback has other manifestations. “The questions of neuroscience are actually questions of engineering,” says Sur, adding that it makes sense for the IITs to invest in neuroscience. Being a product of millions of years of evolution, the brain has perhaps evolved along certain design principles. “It might be that the same principles are important both for engineering and for brains,” says Mitra.
There are particular challenges in India, however. Researchers in the physical sciences and engineering here rarely work with their counterparts in the life sciences. “It's a deep sociological, cultural, intellectual gap that exists. There's a lot to be done,” says Sur. “One centre for computational brain research won't do it all, but it's a start.”