Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?
Get HideMyAss! VPN, PC Mag's Top 10 VPNs of 2016 for 55% off for a Limited Time ×

Submission + - Deep neural networks are easily fooled: Is this Snowcrash for AI? ( 1

anguyen8 writes: Deep neural networks (DNNs) trained with Deep Learning have recently produced mind-blowing results in a variety of pattern-recognition tasks, most notably speech recognition, language translation, and recognizing objects in images, where they now perform at near-human levels. But do they see the same way we do?

Nope. Researchers recently found that it is easy to produce images that are completely unrecognizable to humans, but that DNNs classify with near-certainty as everyday objects. For example, DNNs look at TV static and declare with 99.99% confidence it is a school bus. An evolutionary algorithm produced the synthetic images by generating pictures and selecting for those that a DNN believed to be an object (i.e. “survival of the school-bus-iest”). The resulting computer-generated images look like modern, abstract art. The pictures also help reveal what DNNs learn to care about when recognizing objects (e.g. a school bus is alternating yellow and black lines, but does not need to have a windshield or wheels), shedding light into the inner workings of these DNN black boxes.

Submission + - Deep Neural Networks are Easily Fooled: Is this Snowcrash for AI? ( 1

An anonymous reader writes: A new paper on deep learning produces snowcrash to hack deep neural networks, producing fascinating images and raising security concerns. The paper is called "Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images".

Here is the abstract:

Deep neural networks (DNNs) have recently been achieving state-of-the-art performance on a variety of pattern-recognition tasks, most notably visual classification problems. Given that DNNs are now able to classify objects in images with near-human-level performance, questions naturally arise as to what differences remain between computer and human vision. A recent study revealed that changing an image (e.g. of a lion) in a way imperceptible to humans can cause a DNN to label the image as something else entirely (e.g. mislabeling a lion a library). Here we show a related result: it is easy to produce images that are completely unrecognizable to humans, but that state-of-the-art DNNs believe to be recognizable objects with 99.99% confidence (e.g. labeling with certainty that white noise static is a lion). Specifically, we take convolutional neural networks trained to perform well on either the ImageNet or MNIST datasets and then find images with evolutionary algorithms or gradient ascent that DNNs label with high confidence as belonging to each dataset class. It is possible to produce images totally unrecognizable to human eyes that DNNs believe with near certainty are familiar objects. Our results shed light on interesting differences between human vision and current DNNs, and raise questions about the generality of DNN computer vision.

Submission + - Gates returns to Windows 7 after being unable to install the Windows 8.1 upgrade 3

Artem Tashkinov writes: According to rumors Bill Gate's first day at his office in Redmond turned out to be a complete disaster mixed with ostensibly curse words no one had expected from him. He tried to install the Windows 8.1 upgrade but the updater failed continuously asking to reboot the PC. Microsoft's new C.E.O. Satya Nadella who came to help resolve the situation couldn't sort it out. In the end Gates said he would be returning to Windows 7 for the foreseeable future.

Submission + - Bill Gates Spends First Day Back at MS Failing To Install Windows 8.Reverts to 7 ( 1

JeffClune writes: Bill Gates’s first day at work in the newly created role of technology adviser got off to a rocky start yesterday as the Microsoft founder struggled for hours to install the Windows 8.1 upgrade.

The installation hit a snag early on, sources said, when Mr. Gates repeatedly received an error message informing him that his PC ran into a problem that it could not handle and needed to restart.

After failing to install the upgrade by lunchtime, Mr. Gates summoned the new Microsoft C.E.O. Satya Nadella, who attempted to help him with the installation, but with no success.

While the two men worked behind closed doors, one source described the situation as “tense.”

“Bill is usually a pretty calm guy, so it was weird to hear some of that language coming out of his mouth,” the source said.

A Microsoft spokesman said only that Mr. Gates’s first day in his new job had been “a learning experience” and that, for the immediate future, he would go back to running Windows 7.

Submission + - First Bitcoin ATM arrives in San Diego (

An anonymous reader writes: The first ever Bitcoin ATM has arrived in San Diego. 10News reports that BitcoinATM CEO Evan Rose, a San Diegan, said, "Basically allowing people to buy Bitcoins with cash, cash Bitcoins out right on the spot."

Submission + - Chinese Hackers Infiltrate US Army Database, Compromise Safety Of Thousands Of D ( 1

coolnumbr12 writes: Chinese hackers have infiltrated a sensitive U.S. Army database that contains information about the vulnerabilities of thousands of dams located throughout the United States. TheU.S. Army Corps of Engineers’ National Inventory of Dams (NID) has raised concerns that information gathered in the hack attack could help China carry out a cyberattack on the national electrical power grid.

Submission + - Robotic insect: World's smallest flying robot takes off (

garymortimer writes: Scientists in the US have created a robot the size of a fly that is able to perform the agile manoeuvres of the ubiquitous insects.

This “robo-fly”, built from carbon fibre, weighs a fraction of a gram and has super-fast electronic “muscles” to power its wings.

Its Harvard University developers say tiny robots like theirs may eventually be used in rescue operations.

Submission + - Apple ID Phishing Scam Compromises 100+ Websites (

redletterdave writes: More than 100 different websites had been compromised, but not hacked, to display bogus Apple ID login pages designed to trick Apple users into relinquishing their personal information. All of the compromised sites are hosted by a single IP address registered to an ISP in the Houston area; the majority of these affected sites have not been cleaned, and they continue to present fake, albeit convincing, Apple ID login pages.

Submission + - Bizarre, Squishy Robots Evolve to Run: "Doom Us All" ( 1

semios writes: "A team of researchers has done something remarkable: they've demonstrated simulated evolution in a visible, simple way. Also, they've doomed us all." writes DVICE in response to a viral video of soft robots that have evolved to run. BuzzFeed calls it "the simplest demonstration of evolution you'll ever see." NBC News notes that "simulating creatures has been a staple of biological science for years, but this video shows advances in the field that are both interesting and a little creepy. These virtual 'soft robots' evolved bizarre but somehow natural-looking gaits over thousands of generations." Discover Magazine says "they look ridiculous, and may seem counterintuitive, but these squishy robots get the job done." Evolution was able to design robots with multiple materials akin to muscle, bone and soft tissue. DVICE concludes "So these robots are capable of evolution and could theoretically be printed in a lab, out of structures that could be identical in makeup to those of their human creators. The Cylons are coming, folks. Cute, galloping Cylons. It's just a matter of time. So say we all."
Your Rights Online

Submission + - Police unsure which twin to charge in sexual assaults

An anonymous reader writes: In a real life Prisoner's Dilemma taking place in the French city of Marseille, twin brothers have been arrested for a string of sexual assaults. While say they are sure that one of them committed the crimes (corroborated by a standard DNA test), police were told that it would cost upwards of €1m euros (£850,000, $1.3m USD) to distinguish between them using DNA evidence.

Submission + - How does modularity evolve? ( 1

JimmyQS writes: "As programmers know, modularity is critical to making reusable, adaptable software. However, modularity is not instinctive for beginners and must be learned via painful training. Biology faces a similar problem: modularity is useful to make species more adaptable, but how did it evolve in the first place? Surprisingly, computational simulations of 25,000 generations of evolution reveal that modularity does not evolve because it makes organisms more adaptable. Instead, modularity evolves as a by-product for selection to reduce the "wiring costs" of a network. The discovery greatly advances research into evolving artificially intelligent robots, a field where the inability to evolve modular designs has long been thought to be a key roadblock to evolving truly complex, intelligent neural networks.

The paper was published in the Proceedings of the Royal Society. You can also watch modularity evolve in this video."

Desktops (Apple)

Submission + - Ask Slashdot: Is my Mac infected?

JeffClune writes: "In the last 4-5 months my MacBook Pro (OS X) has become very sluggish, and slows to a crawl while doing some tasks that it used to do fast. My friends complain of the same. I started noticing this slowdown around the same time that reports of widespread Mac malware made the news. How can I know if I am infected? Programs like top/htop do not show any abnormal CPU activity or memory hogs, leaving me to wonder if the malware is hiding itself. Is there a free, trustworthy, effective antivirus/anti-malware program for OS X? Alternately, does OS X get slower over time via bloat? How can I get my Mac's mojo back?!?"

Submission + - Crowdsourced evolution of 3D printable objects (

JimmyQS writes: "The Cornell Creative Machines Lab, which brought us chatbots debating God and unicorns, has developed, a site using evolutionary algorithms and crowdsourcing to design objects that can be 3D printed in materials such as silver, steel or silicone. MIT's Technology Review says "The rules EndlessForms uses to generate objects and their variants resemble those of developmental biology—the study of how DNA instructions unfold to create an entire living organism. The technology is 'very impressive,' says Neri Oxman, director of the MIT Media Lab's Mediated Matter research group. She believes the user-friendliness of the evolutionary approach could help drive the broader adoption of 3-D printing technologies, similar to how easy-to-use image editors fueled the growth of digital photography and graphic manipulation. Oxman [notes] that this could ultimately have an impact on design similar to the impact that blogs and social media have had on journalism, opening the field to the general public." The New Scientist has a quick video tour and describes how the same technology can evolve complex, artificially intelligent brains and bodies for robots that can eventually be 3D printed."

Comment Re:Err... what's the news? (Score 1) 206

Hello. What makes the HyperNEAT approach a breakthrough is its use of a generative encoding based on concepts from developmental biology. Please see my comment one ply deeper in this thread for more information. I encourage you to check out the HyperNEAT publications in order to see why this is very powerful and new technology. Best, Jeff Clune, Postdoctoral Scientist, Michigan State University

Comment Re:What's the news? (Score 1) 206

Hello- I posted this in reply to another comment, but it is relevant to this thread as well. The HyperNEAT technology is actually cutting-edge, and represents a major innovation versus previous neuroevolution techniques. One major thing that differentiates it from previous evolution of ANNs is that HyperNEAT is based on concepts from developmental biology. Specifically, it evolves compositions of geometric coordinate frames that are abstractions of the diffusing chemical gradients of developing embryos. These concepts enable the evolution of regular patterns in neural wiring that have not been seen before in neuroevolution (see, for example, the pictures of evolved brains in my dissertation, which is available at my website: The ability to generate regular wiring patterns enables evolution to search in a small search space of short genomes, yet produce functioning brains with millions or more connections. Of course, this article was written for the popular press, so they did not have the ability to get to this level of detail. For those of you that already know a lot about evolutionary computations and neural nets, I encourage you to read the publications about HyperNEAT, either at my website or at those of other researchers using the technology (e.g., the University of Central Florida). I think you'll then be impressed by the breakthroughs in HyperNEAT. You are correct that evolutionary computation itself has been around for a while. But the science described in this article is pushing that technology further. Best, Jeff Clune Postdoctoral Scientist Michigan State University

Slashdot Top Deals

"Catch a wave and you're sitting on top of the world." - The Beach Boys