Google's "Unsupervised" Self-Learning Neural Network Searches For Cat Pics
June 26, 2012 2:30 PM
comment(s) - last by
(Source: Pop Kitten)
Magic happens at Google Inc.'s (
) X Laboratory, a secret research and development center located at an unknown location in the Bay Area of Northern California. Some projects previously disclosed have been the
augmented reality "Google Goggles"
. Google X is even rumored to be working on
a space elevator
I. AI Neural Net Loves Cats
But one of the most fascinating -- and perhaps frightening -- Google X accomplishment has been its creation of one of the world's largest self-learning
"unsupervised" neural networks
. Consisting of 16,000 computer processors, the array is capable of complex task that are considered impossible using traditional algorithms. One such task is finding cute cats on the internet.
As a test of the nascent cognizant system,
Electrical Engineering Professor
Andrew Y. Ng
and Google fellow Jeff Dean fed the machine 10 million thumbnails of YouTube videos. Without being told exactly what to "look for", the network began to hierarchically arrange data, removing duplicate similar features and group certain images together.
One example was the cat. Thanks to the wealth of cat videos on YouTube, the cyber-brain eventually came to a single dream-like image representing the network's knowledge of what a cat looks like. The network was able to then able to recognize its favorite thing -- cat videos, no matter what subtle variations merry YouTubers come up with to their feline's appearance.
The "cat neuron" holds the learned appearance of what a cat looks like.
[Image Source: Jim Wilson/The New York Times]
The significant part, say researchers, is that the network wasn't told what to look for.
in an interview in
New York Times
, "We never told it during the training, ‘This is a cat.' It basically invented the concept of a cat. We probably have other ones that are side views of cats."
II. Future Systems May Match or Beat Human Brain
Google researchers believe this capability is due to the fact that the network operates similarly to the visual cortex in the human brain. The visual cortex is thought to contain so-called "grandmother neurons", which store key images, such as your love ones' faces. The system developed an idea of what a human face looks like, though it lacked the specificity of known faces stored in the human visual cortex.
The system taught itself what a human looks like. [Image Source: Google]
Dr. Ng describes, "A loose and frankly awful analogy is that our numerical parameters correspond to synapses [in the human brain]."
He says that despite that the network learned what a cat looks like and many basic human features, that it still had far less connections ("synapses") than a human brain. In short, mankind is still winning versus his digital counterpart. Writes the team, "It is worth noting that our network is still tiny compared to the human visual cortex, which is a million times larger in terms of the number of neurons and synapses."
David A. Bader
, executive director of high-performance computing at the
College of Computing, though, says that the team's findings indicate that mankind's era of superiority will be short-lived. He comments, "The scale of modeling the full human visual cortex may be within reach before the end of the decade."
In a difficult test of recognizing 20,000 images, the system performed better than any machine to date, post-learning. The final accuracy was 15.8 percent, 70 percent better than the previous record-holder.
The work was
29th International Conference on Machine Learning
in Edinburgh, Scotland.
The project is now headed out of the top-secret lab and into Google's server farms. Applications that it may be used in include improving results in Google's image search and adaptive speech recognition for Android mobile devices.
But Professor Ng has his sights set on a far more ambitious goal -- a machine that is capable of fully learning, developing into a
fully sentient digital system
. To get there he'll need to wait for the never-ending process of hardware improvements to reach a bit further and he'll also have to work on the fundamental algorithms.
The Google X system is close, but not quite there. He states, "It’d be fantastic if it turns out that all we need to do is take current algorithms and run them bigger, but my gut feeling is that we still don’t quite have the right algorithm yet."
The New York Times
This article is over a month old, voting and posting comments is disabled
Penrose-Hameroff Orch OR Model
6/26/2012 10:06:33 PM
I'm not attaching much to the "cat" thing. It found a series of related patterns but it ends there.
The term cat has so much more information attached to it. Beyond the objective information like anatomy, behavior, history, population, human interaction..etc there is each observers subjective information (countless variables..many working in us right now to make cat seem important).
Treating the human neuron as merely a logic gate multiplied by its synapse complexity is not enough to explain anything remotely close to consciousness. Googles neural net is off by ten to the double digits in orders of magnitude.
Stuart Hameroff and Roger Penrose Orch OR model is nightmare for the AI folks:
RE: Penrose-Hameroff Orch OR Model
6/26/2012 10:25:09 PM
sure the AI is just good enough to recognize cats in 2D space, and will not do much in recognizing cats in 3D space (our real world), but it's a start. Everything starts from the
RE: Penrose-Hameroff Orch OR Model
6/27/2012 12:32:26 PM
This network is modeling the visual cortex of the brain; not the whole brain itself. Extraneous, peripheral information about cats regarding their past history and connection with humans and whatnot is therefore irrelevant in this case, as the system deals with images only - at this stage anyway it would seem. Who can say for sure what'll happen in the future though.
Maybe in a decade we can finally have some proper "agent" type programs that can independently track down information for us in an intelligent manner...
"We don't know how to make a $500 computer that's not a piece of junk." -- Apple CEO Steve Jobs
Report: Wearable Android Google Goggles Will Debut in 2012 for Under $600
February 22, 2012, 11:30 AM
IBM Researchers Develop Neurosynaptic Chips for Cognitive Computing
August 18, 2011, 11:44 AM
Google Developing Self-Piloted Cars
October 11, 2010, 7:22 AM
Leading Neuroscientists Blasts DARPA Cat-Brain Project Calling it a "Scam"
November 24, 2009, 5:18 AM
Carefully Timed Rhythmic Jerks Could Drive Space Elevator
January 6, 2009, 8:15 AM
Apple Says Nude Celebrity Photo Dump Wasn’t Result of iCloud, Find My iPhone Breach
September 2, 2014, 3:26 PM
FCC Orders Advertisers to Cut Out That Racket, Turn Down Commercials
August 29, 2014, 12:49 PM
Dropbox Bows to Competitive Pressure, Provides 1TB of Storage for $10/Month
August 27, 2014, 11:17 AM
Amazon Acquires Twitch for $970 Million
August 25, 2014, 4:37 PM
Facebook Adds Satire Tags to Its Auto-Generated "Related News" Posts
August 18, 2014, 10:43 AM
Comcast, TWC Pull Dinner Gift for FCC Commissioner... Sort Of
August 15, 2014, 1:10 PM
Most Popular Articles
Numerous Leaks Detail 4.7" iPhone 6 Processor, RAM, Cellular and NFC Capabilities
August 29, 2014, 10:37 PM
Apple Builds Not-So-Secret Secret 3-Story Tower for iPhone 6/iWatch Unveil
August 28, 2014, 3:41 PM
Windows 9: "Upgrade Now" Button Coming for Enterprise Updates, ARM Preview in H1 2015
August 26, 2014, 8:00 PM
Buzz Grows Over Upcoming Lenovo Vibe X2's "Layered" Rim, Intel Chip
August 29, 2014, 4:59 PM
Comcast Forced Fees by Reducing Netflix to "VHS-Like Quality"
September 1, 2014, 7:18 PM
Latest Blog Posts
Space Terrorism is a Looming Threat For the United States
Apr 23, 2014, 7:47 PM
Facebook Aims to Provide Internet to "Every Person in the World" with Drones, Satellites
Apr 1, 2014, 10:20 AM
Retail Mobile Sites Experience Outages in Light of Simplexity's Bankruptcy
Mar 14, 2014, 8:48 AM
Tesla vs. BMW: Who Has the Safer EV?
Feb 1, 2014, 2:56 PM
Justice Leaks Details of Next HTC One Two Flagship Phone
Dec 5, 2013, 4:04 PM
More Blog Posts
Copyright 2014 DailyTech LLC. -
Terms, Conditions & Privacy Information