|
SAN JOSE,
Calif.
— In an era when PCs perform like supercomputers, and
supercomputers carry out inhuman feats of calculation,
some of the brightest minds in
Silicon Valley
say there are still crucial ways in which a computer
can't match the problem-solving abilities of our own
brains.
But
Wednesday, at a supercomputing conference in
Portland, Ore.
, a team of scientists from
IBM's
Almaden Research Lab
and several other institutions are planning to announce
two developments that could one day lead to a new kind
of computer — one that uses specially designed
hardware and software to mimic what's inside our heads.
Researchers
from
IBM
and the
Lawrence Berkeley National Laboratory
say they have performed a computer simulation that
matches the scale and complexity of a cat's brain, and
project members from
IBM
and
Stanford
have developed an algorithm for mapping the human brain
at new levels of detail. Eventually, scientists hope
that detailed knowledge will help them build a computer
that replicates the more complex working of a human
brain.
The
developments are early milestones on a long road that
could one day yield applications for business, science
or even the military. Still, veteran computing analyst
Rick Doherty
at the
Envisioneering Group
called the scale and significance of their progress
"jaw-dropping."
The
simulation, for example, did not exactly mimic what a
real cat does in catching a mouse. But it surpassed
earlier efforts that simulated the much simpler brain
structure of a creature the size of a mouse.
Researchers
used an
IBM
supercomputer at the
Lawrence Livermore Lab
to model the movement of data through a structure with 1
billion neurons and 10 trillion synapses, which allowed
them to see how information "percolates"
through a system that's comparable to a feline cerebral
cortex.
The work
is part of a federally funded effort to study what's
known as cognitive computing, starting with what
IBM
project manager
Dharmendra Modha
calls "reverse-engineering the human brain,"
or designing a new computer by first getting a better
understanding of how the brain works.
"The
brain is amazing," said Modha, a computer scientist
who can wax poetic about the capabilities of human gray
matter. "The brain has awe-inspiring capabilities.
It can react or interact with complex, real-world
environments, in a context-dependent way. And yet it
consumes less power than a light bulb and it occupies
less space than a two-liter bottle of soda."
A key
difference between human brains and traditional
computers, Modha says, is that current computers are
designed on a model that differentiates between
processing and storing data, which can lead to a lag in
updating information. The brain works on a more complex
physical structure that can integrate and react to a
constant stream of sights, sounds and other sensory
information.
"The
data can be very ambiguous. When we see a friend's face
in a crowd," Modha said, "she could be wearing
a red sweater or a blue dress, or her hair could be
styled differently, but we're able to get to the
fundamental essence of the pattern and recognize this is
our friend."
Modha
imagines a cognitive computer that could analyze a flood
of constantly updated data from trading floors, banking
institutions and even real estate markets around the
world — sorting through the noise to identify key
trends and their consequences. Or one that could
evaluate pollution, weather and ocean data from
real-time sensors around the world, to monitor global
water supplies.
"As
our digital and physical worlds collide, there is a
tsunami of information," Modha said. "There is
a need for a new kind of intelligence that can sort
through, prioritize and extract the most important
information, much like how the brain deals with sight,
sounds, tastes, touch and smell."
A
cognitive computer might also help soldiers analyze and
react to chaotic events on a battlefield. The research
is the result of a
$5 million
grant from the Pentagon's
Defense Advanced Research Projects Agency
, or
DARPA
, which also funded the forerunner of the Internet. But
like that earlier work, scientists say the study of
cognitive computing could lead in many unexpected
directions.
Stanford
psychology professor
Brian Wandell
, who studies neuroscience, was on the team that
developed a new algorithm for interpreting data from a
kind of noninvasive brain scan. Using supercomputers,
the team has used that data to measure and map the
structure of axons, or thin white threads that help
carry brain signals.
Understanding
these structures could lead to better knowledge of
conditions such as multiple sclerosis or autism, Wandell
said.
"When
you see how something is laid out, you get insights
about how something actually functions," he added.
"So seeing the wiring diagram of the brain will be
helpful for understanding how the brain functions."
|
