Brain not a digital binary machine, understanding its complex functioning can develop better computing model: Kris Gopalakrishnan

About five years ago, Senapathy ‘Kris’ Gopalakrishnan, co-founder of Infosys, made large philanthropic donations towards the niche area of brain research and has continued since to drive interest in this field. One of his latest efforts involved a 30-minute presentation at IIT Bombay’s annual entrepreneurship summit.

Gopalakrishnan explained that there is the clinical aspect of brain research that would help find a cure for or at least arrest the advance of neurodegenerative diseases like Alzheimer’s, Parkinson’s and dementia. And then there is the need to find a better computation model by understanding the brain better.

Nature, the best teacher
“Human history is about mimicking nature. [It is about] wondering why does nature do something and how can we replicate that? Flying is a wonderful example of this. And over time, humans become far better than nature, more capable than nature and also more wasteful than nature,” Gopalakrishnan said during the presentation.

Except, we don’t strictly copy nature, he added. We tend to do things differently. We throw computing power at it. “Brute force,” he called it. “It’s a black box. We can’t explain now how things happen. There are areas that we need to work on,” he said.


According to Gopalakrishnan, most countries today recognise that the marvels of the brain and its workings are the last unknown. ‘What is consciousness?’ is a question we don’t have an answer for yet. “There is a huge opportunity to understand the brain and come out with a better model of computing. We are drowning in software. [Currently,] 70 per cent of budgets are going in maintaining the systems rather than creating new ones. Soon, we will have an explosion of data, and we won’t know what to do with it. That’s why we need a better model of computing,” Gopalakrishnan said.

To make his point, he gave some facts about this powerful organ: “A brain weighs two kilograms and consumes 20 watts of power. A computer with similar power will have 300 megawatts of consumption of power and will probably be a large system, definitely more than two kilograms. That’s why I said that when we want to replicate it [nature], we will throw power at it. And we waste resources,” he said.

Try this to better understand waste of resources. Each time you ask your digital assistant about the weather for the day, large supercomputers are utilised to respond to this seemingly simple query. “If it was done in your home, you may need a grid to just power that computer,” Gopalakrishnan said.

George Boole and John von Neumann, the two fathers of computer science, wondered about the brain. “One person [Boole] said that the brain thinks in binary and Boolean logic. Another person said that the brain has separate areas for processing, storing and registering. Interestingly, we created this whole field of computer science based on these ideas. Surprisingly or not surprisingly, both of them were wrong. The brain is not a digital binary machine. It is analogue,” he said, going back to the history of how the field of computer science evolved, largely by wondering about the brain.

“The new paradigm of computing is you don’t tell the computer how a problem should be solved. You train or teach the computer. You have labelled data. Why do you require it? You need to know what that data pertains to, what it implies, and using this labelled data you train the computer and specify what outcomes are desirable. And the computer then figures out how to go from input to output,” he said, adding that in the new paradigm, you don’t need to know how to solve a problem. If you have input data, you know what desirable outputs are, the machine will figure it out.


Take face recognition, for instance. Where a human can remember 300-400 faces, the machine can remember infinite ones. “Pattern recognition is a problem that people have been trying to solve using finite state automata [where you understand a problem, develop an algorithm, translate that algorithm into your computer program, feed that into a computer, hopefully there are no bugs, the program runs, and solves the problem]. And they were failing because we don’t know how our brain processes patterns,’ Gopalakrishnan said, adding, “Now, machines are even better than humans at recognising faces. You throw computing power at the problem and if you have infinite memory, infinite processing power and a lot of training data, you are able to solve this problem.”

This is what is enabling a new generation of problems to be solved using AI machine learning etc. And that is what is exciting.

And the better we understand the brain, the more efficiently we’ll build the computing systems of tomorrow.