|
"It's going to be a big issue," Geoffrey Hinton, a vice
president with Alphabet Inc's Google, said at a Reuters
Newsmaker event in Toronto on Monday.
Hinton is a pioneer in the booming field of deep learning, which
uses programs known as neural networks to mimic the way humans
learn to perform complex tasks including recognizing images,
sounds and languages.
Hinton led a group of scientists at the University of Toronto
who developed some of the key algorithms that neural networks
use to crunch massive quantities of data, training themselves to
identify patterns so they can mimic the way the human brain
would perform tasks such as driving a car, analyzing potential
financial trades or using medical images to diagnose diseases.
The field has boomed since 2012, when advances in neural
networks enabled Google to add voice recognition to Android
mobile devices and researchers used it to cut error rates in
optical recognition compared with earlier technology, he said.
Neural networks teach themselves to perform complex operations,
making it impossible for their developers to tell government
regulators exactly how those systems work, Hinton said.
"All you need is lots and lots of data and lots of information
about what the right answer is, and you'll be able to train a
big neural net to do what you want," he said.
Deep learning is close to revolutionizing the way certain
diseases are treated. Hinton said neural networks that have
studied millions of medical images will be able to make more
accurate diagnoses than some physicians.
He expects mobile apps to be created that use neural networks to
examine images of skin lesions, advising users when to see a
doctor for a possible biopsy.
"We'd like to make medicine better," Hinton said.
(Reporting by Alastair Sharp in Toronto; Editing by Jim Finkle
and Leslie Adler)
[© 2017 Thomson Reuters. All rights
reserved.] Copyright 2017 Reuters. All rights reserved. This material may not be published,
broadcast, rewritten or redistributed.
 |
|
