i3 - November/December 2016 - 11

By Murray Slovick

Tech

A T EC H TO WATC H

Intelligent Machines
Are the Future

A

COURTESY OF ANKI

rtificial Intelligence (AI) is an area of computer science
that deals with giving machines the ability to model
and/or replicate how natural human intelligence works.
Using "neural networks", a biologically inspired
programming method, a computer can learn from observational
data and "deep learning" a powerful set of techniques by feeding
on large amounts of data. Researchers at
IBM, Microsoft and Google are developing hardware and software that has
shown progress toward visual perception, speech recognition (in particular
understanding what words mean), decision-making, and translation between
languages - making sense of language
the way humans do.
IBM's Watson, designed to defeat
two grandmasters
on the TV quiz show
"Jeopardy" can process
over 500 gigabytes, or
the equivalent of one
million books, every
second. Watson's capabilities focus on Natural
Language Processing, in
other words, written and
spoken text.
Nokia, Panasonic
and Whirlpool recently
discussed how they
are using IBM Watson
Internet of Things technologies. For
example, IBM and Nokia are exploring
opportunities to integrate IBM Watson
IoT with Nokia wearables and smart
devices for home care. Their goal is a
system that helps to detect and alert
caregivers to potential problems such as
deviation from daily routines, abnormal
vital signs and sudden changes in the
home environment. Voice activated
interfaces in the home will be able to
take simple commands (such as "call an
ambulance") and offer reminders to take
C TA . t e c h / i 3

medicines or turn off appliances.
Panasonic's security cameras and
sensors to detect movement, glass
breakage, and door opening will be
coupled with Watson's cognitive computing capabilities. Thanks to video
analytics, a home security system would
know not to react if the neighbors'
children are just fetching their football,
but will automatically alert the police
if an intruder tries to
enter the property.
Using Watson technology a Whirlpool
washing machine will
communicate directly
with a Whirlpool dryer
letting it know what
kind of laundry load to
expect - saving time and
helping to reduce energy
consumption. Using
sensors and cognitive
intelligence, appliances
will learn how people
use them, giving design feedback to
Whirlpool's engineers and offering new
levels of assistance to consumers for reordering detergents and other supplies.

Intelligent APIs
Microsoft's portfolio of machine learning
tools, under the Project Oxford name,
has been rebranded to Microsoft
Cognitive Services. A collection of intelligent APIs (application programming
interfaces) allow solutions to see,
hear, speak, understand and interpret

our needs using natural methods of
communication.
Today, 22 APIs in Cognitive Services can
do things like recognize faces and interpret
natural language even if the app developers are not experts in those fields. The
Emotion API, for example, handles facial
expression in an image. It can detect anger,
contempt, disgust, fear, happiness, neutral,
sadness and surprise.
The visual aspect of the service
can sort through thousands of usersubmitted photos to identify which ones
may contain inappropriate content,
or identify group photos based on a
dominant color scheme.
Microsoft's venture into the world
of AI and deep learning hopes to
implement the technology into its
computer operating systems.

Talking to Machines
Speech is becoming an important
way humans interact with everything
from mobile phones to cars. Allowing
people to easily converse with machines
is a long-standing dream of humancomputer interaction. To build a speech
program that actually sounds human,
a new AI called WaveNet, developed by
Google's DeepMind team, feeds raw audio
waveforms - visual representations of the
shapes sounds take - recorded from real
human speakers to the neural network. In
this way WaveNet mimics human speech
by learning how to form the individual
sound waves a human voice creates.
As well as yielding more naturalsounding speech, using raw waveforms
means that WaveNet can model any kind
of audio, including music.
According to Google, WaveNet reduces
the gap between the state of the art of
speech recognition and human-level
performance by over 50 percent for both
U.S. English and Mandarin Chinese.
n
NOVEMBER/DECEMBER 2016

11
http://CTA.tech/i3
i3 - November/December 2016

Table of Contents for the Digital Edition of i3 - November/December 2016
