AI: Google beats Go
It’s being touted as another milestone on the road to
artificial intelligence: Google’s Deep
Mind computer defeated a master of the ancient Chinese game of Go.
What’s most interesting: not only is the victory a decade
ahead of prognostications, but the neural network trained itself to win. “The
program’s victory marks the rise not merely of the machines but of new
methods of computer programming based on self-training neural networks,”
Spectrum adds. “The program conducted its own research through a
trial-and-error approach that involved playing millions of games against
itself. In this fashion it discovered, one by one, many of the rules of
thumb that textbooks have been imparting to Go students for centuries.”
Energy-friendly Eyeriss chip to boost AI
A new chip designed specifically to implement neural
networks is 10 times as efficient as a mobile GPU, “so it could enable mobile
devices to run powerful artificial-intelligence algorithms locally, rather than
uploading data to the Internet for processing,” MIT reports.
MIT researchers believe neural networks are now “complex,
and are mostly run on high-power GPUs. You can imagine that if you can bring
that functionality to your cell phone or embedded devices, you could still
operate even if you don’t have a WiFi connection. You might also want to
process locally for privacy reasons. Processing it on your phone also avoids
any transmission latency, so that you can react much faster for certain
applications.”
The new chip, “Eyeriss,” has 168 cores, roughly as many as a
mobile GPU. But “the key to Eyeriss’ efficiency is to minimize the frequency
with which cores need to exchange data with distant memory banks, an operation
that consumes a good deal of time and energy. Whereas many of the cores in a
GPU share a single, large memory bank, each of the Eyeriss cores has its own
memory. Moreover, the chip has a circuit that compresses data before sending it
to individual cores.”
Neural network uses “brain-like chips”
Not just a brain-like neural network — but one built from
brain-like plastic memristors rather than silicon.
That’s the work of scientists at the Moscow Institute of Physics and Technology. The chips remember
previous electrical resistance, and so can “effectively work like brain
synapses,” Engadget reports. “They're ideal for creating "true"
neural networks where signal transfers create long-lasting effects.”
Bio-printing: Acoustic tweezers move cells safely
Carnegie Mellon University reports it researchers demonstrated
acoustic tweezers that non-invasively move and manipulate single cells along
three dimensions.
The tool may provide “a promising new method for 3D
bio-printing.”
3D bio-printing may “recreate the complex, multicellular
architecture of biological tissues,” CMU adds. “Researchers have been using a
combination of approaches, but have yet to develop a single method that has the
high level of precision, versatility, multiple dimensionality and single cell
resolution needed to form complex multicellular structures while maintaining
cell viability, integrity and function.”
Why is it needed? “Multicellular structures within living
things are complex and delicate, which makes recreating these structures a
daunting task,” Carnegie says. “For example, the human heart contains more than
2 billion muscle cells. Each of these cells must properly interact with one
another and with their environment to ensure that the heart functions properly.
If those cells aren’t placed correctly, or are damaged, it could potentially
result in any of a variety of heart conditions.”
Flexible transparent pressure sensor
Doctors may someday detect tumors by hand using gloves with
pressure sensors that wrap around and conform to the shape of the fingers while
still accurately measuring pressure distribution, Kurzweil reports.
The transparent, bendable, and sensitive pressure sensor was
developed at the University of Tokyo. It uses carbon nanotubes and graphene,
and is about 8 micrometers thick. It can measure pressure in 144 locations at
once, and “maintain its sensing accuracy even when bent over a radius of 80
micrometers, equivalent to just twice the width of a human hair.”
Smart Wearable senses sweat
Sweat contains dozens of chemical compounds that could
reveal your body’s response to disease, drugs, diet, injury, and stress,
Spectrum notes — and now a wearable sensor can measure levels of specific
molecules in sweat and then wirelessly relay the data to a smartphone.
Developed at the University of California, Berkeley, the
smart wristband combines chemical sensors with a flexible circuit board, and
measures two electrolytes, potassium and sodium ions, and two metabolites,
glucose and lactate. It’s the first to measure more than one analyte at a time.
Handheld microscope can spy cancer cells
With a new handheld microscope, neurosurgeons could
“differentiate cancerous from normal brain tissue at cellular level in real
time in the operating room and determine where to stop cutting,” Kurzweil
reports.
The miniature microscope uses dual-axis confocal microscopy
to illuminate and more clearly see through opaque tissue. It can capture
details up to a half millimeter beneath the tissue surface, where some types of
cancerous cells originate.
Even doctors in their offices could distinguish cancerous
cells, using the new tech developed by engineers at the University of Washington.
Researchers expect to begin testing it as a cancer-screening
tool in clinical settings next year.
Faster online connectivity via airwaves
Are you getting slow and costly connectivity through optical
fiber, cable, and even copper wiring? Soon, wireless access may provide a better
way: Start-up Starry claims it’s deploying the world’s
first millimeter wave band active phased array technology as an alternative to
fixed wireline broadband for consumer internet communications. In other words,
Internet connectivity over the airwaves that is faster than broadband.
Using “underutilized high-frequency spectrum,” the company
plans to offer a nationwide broadband network “capable of delivering internet
speeds of up to one gigabit wirelessly to the home.” Consumers will be able to
install Starry’s products themselves “and connect to the internet in minutes,
without onerous contracts, data caps, or having to wait for an installer.”
The overall costs will be lower, the company adds: “Today it
costs roughly $2,500 per home to deploy traditional broadband service. With
Starry Internet, it will cost about 1/100th of that.”
But if you live in the country, don’t get your hopes up yet:
The service reportedly requires hubs at 1-mile intervals for its line-of-sight transmission.
(Popular
Mechanics notes that “high-frequency waves are very, very bad at going
through things, for example. They can't even penetrate a window. That means in
order for this sort of thing to work, you'd need a device pipes the
internet through your wall into your home and apartment.”)
Starry will launch in the Boston area this summer. Its WiFi
station has a 3.8-inch capacitive touchscreen, and will sell for $350.
Nanowire may improve solar cells
Nanowire chips may be ideal for solar panels, lasers, LEDs,
and other technologies— but making them isn’t easy.
Now scientists at the École
polytechnique fédérale de Lausanne in Switzerland developed a simple and
affordable method to grow the wires by manipulating fluids with microcircuits
on a nanometer scale. (That doesn’t sound too simple…)
This now-standardized method for growing nanowires from
perovskites can be automated on an industrial scale to fabricate chips, the
scientists report.
Concrete clears itself of ice
New conductive concrete carries enough electrical current to
melt ice during winter storms.
Developed at the University
of Nebraska-Lincoln, the concrete has steel shavings and carbon particles
that are warm enough to melt ice but “remain safe to the touch.”
It’s being demonstrated for the Federal Aviation
Administration, and in further testing may be integrated into the tarmac of a
major U.S. airport.
It’s already been tested in a bridge in Lincoln: 52
conductive slabs successfully de-iced its surface during a five-year trial run.
No comments:
Post a Comment