Researchers at IBM have developed a system that stores and processes
information in much the same way that neurons work in the human brain,
auguring the advent of a new generation of smart machines.
Computers have already been taught sensory
capabilities. Now the next step is get them to function like the human
brain. At the International Joint Conference on Neural Networks held in
Dallas, Texas in August, the IBM Cognitive Computing team unveiled a new
type of computer architecture designed to simulate some of the
processes of the human brain. Dubbed
TrueNorth,
this IBM Research project introduces a new programming paradigm
(meaning a fundamental style of computer programming; a way of building
the structure and elements of computer programmes) which includes, among
other elements, an architectural simulator and a library of software
developed for the Systems of Neuromorphic Adaptive Plastic Scalable
Electronics (SyNAPSE) project a year ago.
Distributed processing, Corelet Language
Most modern computer systems are built on what is known as ‘Von
Neumann architecture’ –which creates separate units for storing
information and processing it in a sequential fashion – and they use
programming languages designed specifically for that architecture.
Instead, TrueNorth stores and processes information in a ‘distributed’,
parallel way, like the neurons (processors) and synapses (memory) in a
human brain. The associated software simulates the functioning of a
massive network of neurosynaptic cores with 100 trillion virtual
synapses and two billion neurosynaptic cores. To work with this
approach, an object-oriented Corelet Language has been designed for
creating, composing, and decomposing what are called ‘corelets’ – tiny
neural networks. For example, one corelet can detect motion, while
another can sort images by colour. Individual corelets can be linked
into more and more complex structures. IBM has developed a library of
150 pre-designed corelets to help programmers.
Untangling Big Data
Karlheinz Meier, co-director of the European Union’s Human Brain
Project, explains that non-traditional computing architectures such as
TrueNorth are not intended as a replacement for existing systems.
Inspired by the human brain, and able to sift data fast in order to
perform forecasting, they might be used to solve some problems requiring
big data analysis that the traditional Von Neumann approach cannot
untangle. In fact this system could be applied to many different
domains. For instance, smart sensors equipped with SyNAPSE chips based
on the TrueNorth approach could be used to build special glasses for the
visually impaired, which would collect sensory information and
translate it into a form the wearer could understand. The IBM team is
also working on the possibility of bringing the technology into
smartphones, connected objects and automobiles.
Aucun commentaire:
Enregistrer un commentaire