It's a 'Brain Box'

SpinNaker is an attempt to evolve a computational model that can help understand brain function while solving challenges that lie ahead of chip makers

December 04, 2010 05:08 pm | Updated November 28, 2021 09:19 pm IST

Steve Furber, who designed the ARM chip, now seeks inspiration from biology to evolve a computational model that achieves higher computing and provides a platform for understanding brain function. Photo: G.R.N. Somashekar

Steve Furber, who designed the ARM chip, now seeks inspiration from biology to evolve a computational model that achieves higher computing and provides a platform for understanding brain function. Photo: G.R.N. Somashekar

What brings a celebrated chip maker, who designed the ARM chip that today powers just about every mobile device that we use, to the complex field of neurosciences?

Steve Furber, one of the principle designers of ARM (Acorn RISC machine) architecture, and the man behind the BBC Micro Computer that taught a whole generation of students in the U.K to use computers, today dabbles in a computer of a more organic kind: the brain.

“What does a chip maker have to do with biology, you'll ask,” he says, to a packed hall of students who came to listen to him deliver the first-ever Pinkerton Series lecture at the Infosys campus, organised by the IET, here last week. “The brain has properties that we struggle to achieve. Your brain is a million times more power efficient than the best processor we have.”

They exhibit massively parallelism in processing, are a huge network of neurons and are “astonishingly” power efficient, explains the professor, who carries on his research as ICL Professor of Computer Engineering in the School of Computer Science at the University of Manchester.

Which is why his latest project, the SpinNaker, popularly called the ‘Brain Box', attempts to evolve a computational model that is inspired from the maze of parallel neural networks that are “pinging” inside our brain.

There are two, complementary goals, he explains, in an e-mail interview with The Hindu .

The goal is to use massive parallel computing models to understand brain function, and then, to use that understanding of this ‘energy-efficient' network of neurons in the brain to build more efficient computers. That is, to evolve a computational model, inspired from biology, that can lead to more parallel and fault-tolerant technology.

At the same time, it provides a platform for testing hypotheses that help solve “fundamental mysteries” that still lie at the heart of operation of the brain. Both of these are long-term speculative goals, he adds.

Currently, SpinNaker, the 18-processor chip, is all set to go to fabrication, says Prof. Furber. “We have had SpiNNaker 2-processor test silicon for almost a year. The principal application (of this) is to offer generic support for modelling large-scale systems of spiking neurons in biological real time, to help understand brain function.”

So, what does a SpinNaker look like? The ‘brain box' is a mighty machine comprising over a million ARM processors. The connectivity patterns emulate that of the brain where neurons send messages parallely. The parallelism is important because it is necessary now, more than ever, to find systems that are resilient in an unreliable environment (in terms of component failure).

Modelled in two-dimension, the mesh is a multi-core processor called the SpinNaker chip, that comprises 20 processing cores. It establishes parallelism by connecting to its local peers through a network that provides high bandwidth communication. The interconnection pattern is meant to evolve, similar to how a brain learns.

Troubles ahead

But why build a system this massive, when existing chip sets are evolving at a consistent rate using a simpler systems-on-chip model? Prof. Furber points out that there is trouble ahead for the chip makers, who have to constantly up the ante on energy-efficiency, even as transistors must keep shrinking.

With each shrinking transistor, we reach atomic scales, making the material and component less predictable or reliable, he explains. This will be a huge problem for the industry in the coming years. Understanding parallel programming will also be a challenge.

He is less optimistic about multi-core processors, which he feels are currently more of a “cut-and-paste job”. “So if you're taking home one of those multi-core machines, the chances are that you won't be using more than one of them effectively at a given time,” he explains.

Unviable for smaller players

Another significant issue he brings up is that of cost. He points out that the fabless semiconductor start-up model, that drove innovation in chip design in the 1990s, is no longer economically feasible. “This has put the major innovation back in the hands of the large, established companies.”

Given that Acorn, the technology firm that built the iconic ARM, was a small player (that later tied up with Apple to form a joint-venture firm ARM Limited), this is a “significant concern” for Prof. Furber.

He observes that this is seeing the end of the “entrepreneurial dream” of the past, where small players can ride in on the power of their innovation. “That's almost ended in the U.K.,” he adds.

0 / 0
Sign in to unlock member-only benefits!
  • Access 10 free stories every month
  • Save stories to read later
  • Access to comment on every story
  • Sign-up/manage your newsletter subscriptions with a single click
  • Get notified by email for early access to discounts & offers on our products
Sign in

Comments

Comments have to be in English, and in full sentences. They cannot be abusive or personal. Please abide by our community guidelines for posting your comments.

We have migrated to a new commenting platform. If you are already a registered user of The Hindu and logged in, you may continue to engage with our articles. If you do not have an account please register and login to post comments. Users can access their older comments by logging into their accounts on Vuukle.