Rise of the machines: Google AI experiment may lead to robots that can learn WITHOUT human input
Rise of the machines: Google AI experiment may lead to
robots that can learn WITHOUT human input
Generative Adversarial Networks create digital content
based on real-life
Google project pits AI algorithms against each other to
refine this output
The results could one day lead to machines that can learn
without human input
By TIM COLLINS FOR MAILONLINE PUBLISHED: 07:13 EDT, 18
April 2017 | UPDATED: 09:21 EDT, 18 April 2017
Machines that can think for themselves - and perhaps turn
on their creators as a result - have long been a fascination of science
fiction.
And creating robots that can learn without any input from
humans is moving ever closer, thanks to the latest developments in artificial
intelligence.
One such project seeks to pit the wits of two AI algorithms
against each other, with results that
could one day lead to the emergence of such intelligent machines.
BATTLE OF THE BOTS
Google's Generative Adversarial Network works by pitting
two algorithms against each other, in an attempt to create convincing
representations of the real world.
These 'imagined' digital creations - which can take the
form of images, videos, sounds and other content - are based on data fed to the
system.
One AI bot creates new content based upon what it has
been taught, while a second critiques these creations - pointing out
imperfections and inaccuracies.
And the process could one day allow robots to learn new
information without any input from people.
Researchers at the Google Brain AI lab have developed a
system known as a Generative Adversarial Network (GAN).
Conventional AI uses input to 'teach' an algorithm about
a particular subject by feeding it massive amounts of information.
This knowledge can then be employed for a specific task -
facial recognition being just one example.
GANs seek to generate new content from this learned
information, creating digital content like pictures and video based on their
understanding of similar real life images and footage.
Google's approach is to set two algorithms against each
other, to further refine these 'imaginings'.
One AI bot creates new content based upon what it has
been taught about the real world, while a second critiques these creations -
pointing out imperfections and inaccuracies.
This allows the system to create more realistic images,
sounds and other original creations that are far more realistic than if the
first bot was working alone.
And the process could one day allow robots to learn new
information without any input from people - a process called 'unsupervised
learning' that would represent a giant leap forward in AI technology.
Speaking to Wired, Dr Ian Goodfellow, who works at Google
mind, said: 'If an AI can imagine the world in realistic detail—learn how to
imagine realistic images and realistic sounds—this encourages the AI to learn
about the structure of the world that actually exists.
'You can think of this like an artist and an art critic.
'The generative model wants to fool the art critic—trick
the art critic into thinking the images it generates are real.'
Artificial intelligence systems rely on neural networks,
which try to simulate the way the brain works in order to learn.
MACHINE LEARNING
Artificial intelligence systems rely on neural networks,
which try to simulate the way the brain works in order to learn.
These networks can be trained to recognise patterns in
information - including speech, text data, or visual images - and are the basis
for a large number of the developments in AI over recent years.
They use input from the digital world to learn, with
practical applications like Google's language translation services, Facebook's
facial recognition software and Snapchat's image altering live filters.
But the process of inputting this data can be extremely
time consuming, and is limited to one type of knowledge.
These networks can be trained to recognise patterns in
information - including speech, text data, or visual images - and are the basis
for a large number of the developments in AI over recent years.
They use input from the digital world to learn, with
practical applications like Google's language translation services, Facebook's
facial recognition software and Snapchat's image altering live filters.
But the process of inputting this data can be extremely
time consuming, and is limited to one type of knowledge.
This is not the first time that Google has set AI bots
against each other, to expand the limits of this type of machine learning.
In February, a Google team used a game they designed to
examine whether competing algorithms would work together or turn on each other.
These experiments showed that AI may be more or less
likely to work together depending on the situation.
The results could add to our understanding and control of
complex multi-agent systems such as the economy, traffic systems, or the
ecological health of our planet – all of which depend on our continued
cooperation.
Comments
Post a Comment