MIT researchers create "brain-on-a-chip"

The new neuromorphic chip aims to make tech work more like the human brain and is capable of processing significantly more information faster

10Oct

In a new research journal, scientists from the Massachusetts Institute of Technology (MIT) have unveiled a new chip which will be capable of processing facts, patterns, images and video at lightning speeds, it has been revealed. The chip is being hailed as the next leap forward in the AI and machine learning (ML) field.

In order to do this, the MIT scientists have explored a cutting-edge area of science called neuromorphic computing. This arm of computing is based on a concept first put forward by engineer Carver Mead in the 1980's and looks to mimic neurobiological structures to develop electronic circuits.

The MIT researchers have created what they call a "brain on a chip". Instead of using the standard binary on/off signaling most commercial computers use, these neuromorphic chips work like neurons in the brain, exchanging burst of electricity at varying intensity in a much more analog way.


Visit Innovation Enterprise's Machine Learning Innovation Summit, part of the DATAx New York festival, on December 12–13, 2018


Following the announcement, in a statement to CNBC, MIT's Research Laboratory of Electronics and Microsystems Technology Laboratories lead researcher, Jeehwan Kim said: "Supercomputer-based artificial neural network operation is very precise and very efficient. However, it consumes a lot of power and requires a large footprint."

However, as the new architecture mimics the human brain, it uses significantly less power to carry out ML processes, with small chips capable of using up to 1,000 times less energy while performing at large-bank supercomputer level.

These advancements could supercharge every area of computing from robotics to autonomous vehicles and is being considered as one of the solutions to remedy the impending end of Moore's Law.

"These new kinds of chips should increase dramatically the use of ML," reported a Deloitte market analysis in early 2018. "This will enable applications to consume less power and at the same time become more responsive, flexible and capable," it added.

Google drops out of  10bn pentagon deal small

Read next:

Google drops out of $10bn Pentagon deal

i