learning. It also features isolated pipelines for computation and data management, as well as High Bandwidth Memory (HBM) to accelerate data movement.
Nervana, which Intel expects to introduce during the first half of this year, will
“deliver sustained performance near
theoretical maximum throughput,” he
adds. It also includes 12 bidirectional
high-bandwidth links, enabling multiple interconnected engines for seamless scalability, a key requirement for
increased performance through scale.
Into the Future
An intriguing aspect of emerging chip
designs for AI, deep learning, and
machine learning is the fact that
low-precision chip designs increasingly prevail. In many cases, reduced-precision processors conform better to
neuromorphic compute platforms and
accelerate the deployment and possibly
training of deep learning algorithms.
Simply put: they can produce similar
results while consuming less power,
in some cases by a factor of 100. While
algorithms running on today’s digital
processors require high numerical precision, the same algorithms operating
on low precision chips in a neural net
excel, because these systems adapt dynamically by examining data in a more
relational and contextual way (and are
less sensitive to rounding errors).
This makes the technology perfect
for an array of machine learning tasks
and technologies, including drones;
automated vehicles; intelligent per-
sonal assistants such as Amazon’s
Alexa, Microsoft’s Cortana, or Apple’s
Siri; photo and image recognition
systems, and search engines, includ-
ing general services like Bing and
Google but also those used by retail-
ers, online travel agencies, and others.
It also supports advanced functionality
like real-time speech-to-text transcrip-
tion and language translations.
In the end, says Gregory Diamos, a
senior researcher at Baidu, specialized
machine learning chips have the potential to change the stakes and usher in
an era of even greater breakthroughs.
“Machine learning has already made
tremendous progress,” he says. “
Specialized chips and systems will continue to close the gap between computers
and human performance.”
Caulfield, A., Chung, E., Putnam, A., Angepat,
H., Fowers, J., Haselman, M., Heil, S., Humphrey,
M., Kaur, P., Kim, J. Y., Lo, D., Massengill, T.,
Ovtcharov, K., Papamichael, M., Woods, L.,
Lanka, S., Chiou, D., and Burger, D.
A Cloud-Scale Acceleration Architecture,
October 15, 2016. Proceedings of the
49th Annual IEEE/ACM International
Symposium on Microarchitecture, IEEE
Computer Society. https://www.microsoft.
Samel, B., Mahajan, S., and Ingole, A.M.
GPU Computing and Its Applications,
International Research Journal of
Engineering and Technology (IRJE T).
Volume: 03 Issue: 04, Apr-2016.
Shafiee, A., Nag, A., Muralimanohar, N.,
Balasubramonian, R., Strachan, J.P., Hu, M.,
Williams, S.R., and Srikumar, V.
ISAAC: A Convolutional Neural Network
Accelerator with In-Situ Analog Arithmetic
in Crossbars, 2016 ACM/IEEE 43rd Annual
International Symposium on Computer
Architecture (ISCA), pp. 14-26, 2016, ISSN
Shirahata, K., Tomita, Y., and Ike, A.
Memory reduction method for deep
neural network training, 2016 IEEE
26th International Workshop on
Machine Learning for Signal Processing
(MLSP), 2016, pp. 1-6. doi: 10.1109/
Samuel Greengard is an author and journalist based in
West Linn, OR.
© 2017 ACM 0001-0782/17/05 $15.00
reached a point
can deploy models
expertise about the
MATH AND LANGUAGE
science at the
University of Michigan, studied
computer science as an
undergraduate at the Technical
University of Sofia, “I was
interested in math and
languages; French, Russian, and
English. I was not sure how to
combine those two interests,”
Radev explains. “When the first
personal computers came
around, I thought it would be a
good way to combine my
Radev completed his
undergraduate degree at the
University of Maine at Orono,
before going on to earn a
Ph. D. in computer science at
Columbia University in New
York in 1999 (while serving as
an adjunct assistant professor
in the department of computer
science). His focus was on
natural language processing
and computational linguistics,
working on algorithms to teach
human languages to computers.
Even before graduating,
Radev was hired by IBM in 1998
to work on the team that built
the first question/answer system
at the company’s Thomas J.
Watson Research Center in
Hawthorne, N Y. “After a year-
and-a-half at IBM, I started at
the University of Michigan” in
January 2000, he adds, “and I
have been there since.”
Radev now is involved
with building spoken dialog
systems for student advising,
and he serves on the executive
committee at the Association for
Computational Linguistics, an
organization for those working
on problems involving natural
language and computation.
He has also served as co-chair of the North American
Olympiad (NACLO), in which
thousands of high school
students in the U, S, and Canada
compete to solve problems in
natural language processing
and computational linguistics.