Super robots to be given eyes & intelligence to identify terrorists and enemy aircraft
AI BREAKTHROUGH: Super robots to be given BRAINS to
identify terrorists and enemy aircraft
SUPER-ROBOTS are to be given 'brains' which mimic human
eyes to enable them to identify terror suspects in a crowd, enemy aircraft or
find missing children.
By ALIX CULBERTSON
PUBLISHED: 00:00, Fri, Aug 4, 2017 | UPDATED: 21:51, Fri,
Aug 4, 2017
Robots are set to get eyes which can spot terrorists
Scientists are "well within reach" of creating
an artificial retina which could be used inside advanced image recognition
cameras to help nations' spies and police with investigations.
The exciting breakthrough has been likened to Star Trek's
Commander Data as it could contain neuristors - neurons and transistors - and
multi-circuit components which emulate the speed at which the human brain
works.
Researchers at the Georgia Institute of Technology and
nearby Binghamton University have been awarded a £5.4million grant by the US
Department of Defence to develop the technology.
They will be expanding on current findings in which new
metal oxide materials buzz electronically at the nanoscale which emulate the
way human brain networks buzz on a cellular level.
"Brain-mimicking" will be
"significantly" advanced by the new findings and will help
scientists' currently limited understanding of how the human brain works.
The new technology consists of ultra-compact circuits
called neuristors which sense light, compute an image from it then store the
image.
Unlike previous attempts, the new findings mean all three
of the functions will occur simultaneously and nearly instantaneously - just
like in a human brain.
Alan Doolittle, a professor at Georgia Tech's School of
Electrical and Computer Engineering, said: "The same device senses,
computes and stores the image.
"The device is the sensor, and it's the processor,
and it's the memory all at the same time."
Despite computers being extremely rapid in 2017, there is
still a processing and memory lag time as machines need to tap into separate
memory components, then process, in a never-ending process.
Professor Doolittle, added: "That back-and-forth
from memory to microprocessor has created a bottleneck."
With the new retina, signals will flow "very freely,
more like they do through the brain" which then massively expands the
number of possible pathways it can compute, which in turn helps it compute
"powerfully and swiftly".
Thousands of photos will be able to be stored in the new
devices to allow them to match up what they see with previous images they have
seen.
The Georgia Tech Horizons research magazine, said: "The
retina could pinpoint known terror suspects in a crowd, find missing children,
or identify enemy aircraft virtually instantaneously, without having to trawl
databases to correctly identify what is in the images.
"Even if you take away the optics, the new neuristor
arrays still advance artificial intelligence."
There is also the possibility of the retina learning to
select further information out of the images it sees.
Professional Doolittle, added: "It will work with
anything that has a repetitive pattern like radar signatures, for example.
"Right now, that's too challenging to compute,
because radar information is flying out at such a high data rate that no
computer can even think about keeping up."
The development is a big step towards mimicking the human
brain, although Professor Doolittle admitted there is still a long way to go.
He said: "We're not going to reach circuit
complexities of that magnitude, not even a tenth.
"Also, currently science doesn't really know yet
very well how the human brain works, so we can't duplicate it."
Comments
Post a Comment