SoylentNews

takyon writes:

Optalysys claims to have implemented a convolutional neural network (CNN) using its optical supercomputer/coprocessor:

Optalysys says they have built the world's first implementation of a convolutional neural network using its optical coprocessor.

The UK-based company has developed optical computing hardware that uses lasers and spatial light modulators (SLMs) to perform complex numerical processing at extremely high speeds and using very little power. The technology has been implemented as an HPC coprocessor, which is meant to be attached to a conventional computer – a desktop system or a server. The company's first prototype was announced in 2014.

Although not much detail has been provided on this latest application of the technology, Optalysys says they have implemented a CNN based on the MNIST dataset of hand-drawn numerals. The dataset is comprised of 60,000 training characters and 10,000 testing characters. According to the company, its optical laser technology enables them to process this model several orders of magnitude faster than conventional electronic hardware and does so at a fraction of the energy consumption.

[...] Optalysys is one of a growing number of startups that is applying optical computing to AI. Those companies include Lighton, Light Intelligence, Fathom Computing, and Lightmatter. (We covered that latter two here and here.) All are using various forms of optical technology to encode neural networks and are promising huge speedups and much better energy efficiency than traditional CPU/GPU-based machine learning.

MNIST database. Press release.

Previously: Computing With Lasers Could Power Up Genomics and AI
Optalysys, Back in the (Press Release) News

Original Submission

aristarchus writes:

From the Vox:

Many American pundits seem to firmly believe that the country stands at a precipice in which young, left-wing college students and recent graduates are the leading edge of a rising tide of illiberalism that comes in the form of “political correctness” and poses a clear and present danger to free speech and rational discourse.

[...] The alarm about student protesters, in other words, though not always mistaken about particular cases, is generally grounded in a completely mistaken view of the big-picture state of American society and public opinion, both on and off campus.

[...] Since the 1970s, the General Social Survey has posed a question about whether five hypothetical speakers should be allowed to give a speech in your community — a communist, a homosexual, an opponent of all religion, a racist, and a person who favors replacing the elected government with a military coup.

Justin Murphy of the University of Southampton in the United Kingdom aggregated trend data about all five kinds of speakers and found that public support for free expression has been generally rising

Original Submission

Fnord666 writes:

"Leaky etcd servers could be a boon to data thieves and ransomware scammers."

etcd is described as "A distributed, reliable key-value store for the most critical data of a distributed system.".

Thousands of servers operated by businesses and other organizations are openly sharing credentials that may allow anyone on the Internet to log in and read or modify potentially sensitive data stored online.

In a blog post published late last week, researcher Giovanni Collazo said a quick query on the Shodan search engine returned almost 2,300 Internet-exposed servers running etcd, a type of database that computing clusters and other types of networks use to store and distribute passwords and configuration settings needed by various servers and applications. etcd comes with a programming interface that responds to simple queries that by default return administrative login credentials without first requiring authentication. The passwords, encryption keys, and other forms of credentials are used to access MySQL and PostgreSQL databases, content management systems, and other types of production servers.

Maybe it's just me, but if the phrases "store for the most critical data of a distributed system" and "Internet facing" both occur in your description of a node of your architecture, you're probably doing it wrong.

Original Submission

An Anonymous Coward informs us of a link:

Sex Workers Say Porn on Google Drive Is Suddenly Disappearing

[...] Six porn performers I talked to and more on social media said that they suddenly can't download adult content they keep on Google Drive. They also said they can't a[sic] share that content with other accounts or send to clients. In some cases, the adult content is disappearing from Drive without warning or explanation. The porn performers I talked to started sounding the alarm on Twitter last week. They said that Google Drive no longer seemed sex-trade friendly, detailing error messages and sharing cloud storage alternatives with each other.

When I asked about sexual content being blocked on Drive, a spokesperson for Google directed me to the Drive policy page—specifically the section on sexually explicit material, which says, "Do not publish sexually explicit or pornographic images or videos.... Additionally, we do not allow content that drives traffic to commercial pornography." Writing about porn and sex is permitted, the policy states, as long as it's not accompanied by sexually explicit images or videos. According to Google, Drive uses a combination of automated systems and manual review to decide what's in violation.

[...] "It seems like all of our videos in Google Drive are getting flagged by some sort of automated system," Stone said. "We're not even really getting notified of it, the only way we really found out was one of our customers told us he couldn't view or download the video we sent him."

Stone's files aren't removed from Drive, but when she tries to play the video or download it, she said Google gives her an error message: "Whoops! There was a problem playing this video" with an option to download the item, but the download link doesn't work.

Some sex workers are wondering if this has something to do with the impending vote on the SESTA-FOSTA bill, which is on the Senate floor for debate this week. [ed. note: it was passed]

It could also be that Google is suddenly enforcing its Terms of Service without warning.

[...] "I don't believe that Google should be allowed to dictate what you and another consenting adult send to each other through email."

Original Submission

Fnord666 writes:

YouTube expands firearms restrictions, more gun videos to be banned

"Some gun-related channels are already feeling the heat."

YouTube is placing more restrictions on weapons-related videos, focusing on guns with new, forthcoming policy changes. According to a Bloomberg report, YouTube intends to ban videos that "promote or link to websites selling firearms and accessories," including bump stocks, beginning this April. The new policy will also prohibit instructional videos that detail how to build firearms.

These restrictions come over a month after the school shooting in Parkland, Florida and just a few days before the March for Our Lives rally organized by the student survivors of the Parkland shooting. YouTube took similar action after the Las Vegas shooting last year by banning gun-modification tutorials.

"We routinely make updates and adjustments to our enforcement guidelines across all of our policies," a YouTube representative said in a statement to Bloomberg. "While we've long prohibited the sale of firearms, we recently notified creators of updates we will be making around content promoting the sale or manufacture of firearms and their accessories."

[...] While some may see YouTube's new firearms policy as ambiguously worded, it's the forthcoming implementation that will get the most reaction from firearms channels. Plenty of YouTubers have seen their content demonetized or removed due to the way YouTube's algorithm and moderators filter out potentially offensive content and content that goes against Community Guidelines. It's possible that gun-related videos that do not explicitly violate the new rules will get caught up in the first rounds of YouTube's upcoming purge.

[...] With the upcoming policy, YouTube will join the bevy of other companies, including Dick's Sporting Goods and Walmart, that have instituted new restrictions on the promotion or sales of firearms in the wake of the Parkland shooting.

Gun videos migrate to porn sites as YouTube cracks down

An Anonymous Coward inform us of a link:

THERE is a bunch of unusual videos turning up on porn streaming sites as America's gun advocates cry foul.

YOUTUBE is going to start banning videos related to the sale or manufacture of guns next month, so as a way to make up for it, firearm aficionados are jumping ship to Pornhub — where they can post pretty much any clip they'd like.

Gun videos migrate to porn sites as YouTube cracks down

Original Submission #1 Original Submission #2

An Anonymous Coward writes:

Richard Dawkins is responding to what he called the "stirring towards atheism" in some Islamic countries with a programme to make free downloads of his books available in Arabic, Urdu, Farsi and Indonesian.

The scientist and atheist said he was "greatly encouraged" to learn that the unofficial Arabic pdf of the book had been downloaded 13m times. Dawkins writes in The God Delusion about his wish that the "open-minded people" who read it will "break free of the vice of religion altogether". It has sold 3.3m copies worldwide since it was published in 2006 – far fewer than the number of Arabic copies that Dawkins believes to have been downloaded illegally.

Richard Dawkins to give away copies of The God Delusion in Islamic countries

Original Submission

Fnord666 writes:

"New designs flip between two metastable states to generate force, pack themselves."

I have to admit that even as someone who is fascinated by most insects, the earwig freaks me out. Upon seeing one, I'm typically too busy trying to squash it to notice any details about its anatomy. So it was a bit of a surprise to find out that not only do they have wings, but their wings are world record holders in a specific aspect of insect winginess: they take up the least space when folded compared to their extended size. The ratio between these states can reach as high as 18-to-one.

With that fact in mind, I was less surprised to find out that researchers have decided to study this bit of biology to see if they can derive any insights from what evolution has done with the earwig. In today's issue of Science, there is a report on what has been learned by three researchers: Jakob Faber and André Studart of ETH Zurich and Andres Arrieta of Purdue University. They find that, to mimic the earwig's wing, an origami-style folding approach won't do. Instead, they have designed and 3D-printed a selection of meta-stable designs that, with a small input of energy, rapidly flip between folded and unfolded states.

When many people, including most materials scientists, think of folding, their first thought is origami. But the research team found that the earwig's "exquisite natural folding system" behaves in a way that "cannot be sufficiently described by current origami models." Part of the issue is one of materials science: there are certain folding patterns in the wing that just can't be done by creating a crease in a single material or using the straight lines of origami. In addition, the wing is bi-stable, holding itself in place during flight with minimal input from muscles and folding up entirely without any muscular energy being expended.

The secret to this is partly that biology is not limited to either straight lines or a single material. In fact, the joints on an earwig's wing are rich in a protein called resilin, which forms a flexible polymer that can store and release energy as it is bent and relaxed.

Original Submission

martyb writes:

Don't try to ignore past failures, learn from it and move on:

Insights from past failures can help boost performance on a new task -- and a new study is the first to explain why. US researchers report that writing critically about past setbacks leads to lower levels of the "stress" hormone, cortisol, and more careful choices when faced with a new stressful task, resulting in improved performance. The study, published today in open access journal Frontiers in Behavioral Neuroscience, is the first demonstration that writing and thinking deeply about a past failure improves the body's response to stress and enhances performance on a new task. This technique may be useful in improving performance in many areas, including therapeutic settings, education and sports.

[...] But why does this counter-intuitive approach lead to benefits? To investigate this question, Brynne DiMenichi, a doctoral candidate from Rutgers University-Newark, together with other researchers from the University of Pennsylvania and Duke University, examined the effect of writing about past failures on future task performance in two groups of volunteers.

A test group wrote about their past failures while a control group wrote about a topic not related to themselves. The researchers used salivary cortisol levels to provide a physiological readout of the stress experienced by the people in both groups. These levels were comparable across the test and control groups at the start of the study.

DiMenichi and colleagues then measured the performance of the volunteers on a new stressful task and continued to monitor their cortisol levels. They found that the test group had lower cortisol levels compared to the control group when performing the new challenge.

"We didn't find that writing itself had a direct relationship on the body's stress responses," says DiMenichi. "Instead, our results suggest that, in a future stressful situation, having previously written about a past failure causes the body's stress response to look more similar to someone who isn't exposed to stress at all."

The researchers also found that volunteers who wrote about a past failure made more careful choices on a new task, and overall performed better than the control group.

"Together, these findings indicate that writing and thinking critically about a past failure can prepare an individual both physiologically and cognitively for new challenges," observes DiMenichi.

Journal Reference: Brynne C. DiMenichi, Karolina M. Lempert, Christina Bejjani, Elizabeth Tricomi. Writing About Past Failures Attenuates Cortisol Responses and Sustained Attention Deficits Following Psychosocial Stress. Frontiers in Behavioral Neuroscience, 2018; 12 DOI: 10.3389/fnbeh.2018.00045

Original Submission

canopic jug writes:

Bunnie Huang, hardware hacker, wrote a brief article about transparency versus liability in the context of open hardware. He covers some of the tradeoffs without going into depth.

[...] Should a buggy library you develop be used in a home automation appliance that later causes a house to catch fire, you get to walk away scot-free, thanks to the expansive limited-liability clauses that are baked into every open source software licence.

Unfortunately, hardware makers don't get to enjoy that same luxury. Beyond guaranteeing a product free from workmanship or material defects, consumer protection law often requires an implied or express 'fitness for purpose' guarantee – that a piece of hardware is capable of doing what it's advertised to do. The latest controversy over Spectre/Meltdown indicates that more people than not feel CPU makers like Intel should be liable for these bugs, under the 'fitness for purpose' theory.

Open hardware makers should be deeply concerned. [...]

At BlackHat 2014, Dan was more specific regarding software and raised, with Poul-Henning Kamp, the idea that normal liability laws should also apply to software. But with that liability in place, exemptions should be available if vendors supply complete and buildable source code along with a license that allows disabling any functionality or code that the licensee decides against. Poul-Henning has called for a long time for changes to liability laws for software.

Original Submission

canopic jug writes:

Washington's governor signs anti-censorship bill for student journalists into law (archive) (alt). Similar state legislation has been attempted several times in the past.

Gov. Jay Inslee signed Senate Bill 5064 Wednesday in front of a group of students, teachers and school administrators in Olympia. The new law, which goes into effect this June, makes Washington the last state on the West Coast to pass an "anti-Hazelwood law," a reference to a 1988 U.S. Supreme Court ruling that made it legal for school administrators to censor content in school newspapers and other student-run media.

[...] [Senator Joe Fain, R-Auburn], sponsored SB 5064 this year, although the legislation has been introduced in Olympia four different times, in various forms, by three different lawmakers since 2007.

So the new law gives a big boost to student newspapers by preventing school administrators from censoring content. Now the hair-splitting will begin regarding what defines illegal content or harassement.

Original Submission

takyon writes:

A new version of the NEST algorithm could dramatically reduce the amount of memory required to run a whole human brain simulation, while increasing simulation speed on current supercomputers:

During the simulation, a neuron's action potentials (short electric pulses) first need to be sent to all 100,000 or so small computers, called nodes, each equipped with a number of processors doing the actual calculations. Each node then checks which of all these pulses are relevant for the virtual neurons that exist on this node.

That process requires one bit of information per processor for every neuron in the whole network. For a network of one billion neurons, a large part of the memory in each node is consumed by this single bit of information per neuron. Of course, the amount of computer memory required per processor for these extra bits per neuron increases with the size of the neuronal network. To go beyond the 1 percent and simulate the entire human brain would require the memory available to each processor to be 100 times larger than in today's supercomputers.

In future exascale computers, such as the post-K computer planned in Kobe and JUWELS at Jülich in Germany, the number of processors per compute node will increase, but the memory per processor and the number of compute nodes will stay the same.

Achieving whole-brain simulation on future exascale supercomputers. That's where the next-generation NEST algorithm comes in. At the beginning of the simulation, the new NEST algorithm will allow the nodes to exchange information about what data on neuronal activity needs to [be] sent and to where. Once this knowledge is available, the exchange of data between nodes can be organized such that a given node only receives the information it actually requires. That will eliminate the need for the additional bit for each neuron in the network.

With memory consumption under control, simulation speed will then become the main focus. For example, a large simulation of 0.52 billion neurons connected by 5.8 trillion synapses running on the supercomputer JUQUEEN in Jülich previously required 28.5 minutes to compute one second of biological time. With the improved algorithm, the time will be reduced to just 5.2 minutes, the researchers calculate.

Also at the Human Brain Project.

Extremely Scalable Spiking Neuronal Network Simulation Code: From Laptops to Exascale Computers (open, DOI: 10.3389/fninf.2018.00002) (DX)

Previously: Largest neuronal network simulation achieved using K computer (2013)

Original Submission

takyon writes:

Physicist Proposes Alternative to Black Holes

A physicist has incorporated a quantum mechanical idea with general relativity to arrive at a new alternative to black hole singularities. What do you get when you cross two hypothetical alternatives to black holes? A self-consistent semiclassical relativistic star, according to Raúl Carballo-Rubio (International School for Advanced Studies, Trieste, Italy) whose recently published results in the February 6th Physical Review Letters [DOI: 10.1103/PhysRevLett.120.061102] [DX] describe a new mathematical model for the fate of massive stars.

When a massive star comes to the end of its life, it goes supernova, leaving behind a dense core that — according to conventional thought — continues to collapse to form either a neutron star or black hole. To which fate a particular star is destined comes down to its mass. Neutron stars find a balance between the repulsive force of quantum mechanical degeneracy pressure and the attractive force of gravity, while more massive cores collapse into black holes, unable to fight the overwhelming pull of their own gravity.

Now, Carballo-Rubio adds an extra force into the mix: quantum fluctuations. Quantum mechanics has shown that virtual particles spontaneously pop into and out of existence — the effects can be measured best in a vacuum, but these fluctuations can happen anywhere in spacetime. These particles can be thought of as fluctuations of positive and negative energy that under normal conditions would cancel out. But the extreme gravity of compact objects breaks this balance, effectively generating negative energy. This negative energy creates a repulsive gravitational force. "The existence of quantum [fluctuations] due to gravitational fields has been known since the late 1970s," explains Carballo-Rubio. But physicists didn't know how to take this effect into account in collapsing stars.

Carballo-Rubio derived equations that combine general relativity and quantum mechanics in a way that accounts for quantum fluctuations. Moreover, he found solutions that balance attractive and negative gravity for stellar masses that would otherwise have produced black holes. Dubbing them "semiclassical relativistic stars," these compact objects do not fully collapse under their own weight to form an event horizon, and are therefore not black holes.

Original Submission

takyon writes:

Epic Games' Tim Sweeney on creating believable digital humans

Epic Games stunned everyone a couple of years ago with the realistic digital human character Senua, from the video game Hellblade. And today, the maker of the Unreal Engine game tools showed another astounding demo, dubbed Siren, with even more realistic graphics.

CEO Tim Sweeney said technologies for creating digital humans — from partners such as Cubic Motion and 3Lateral — are racing ahead to the point where we won't be able to tell the real from the artificial in video games and other real-time content.

[...] [Kim Libreri:] The other big thing for us, you may have seen the Microsoft announcements about their new raytracing capabilities in DirectX, DXR. We've partnered with Nvidia, who have the new RTX raytracing system, and we thought about how to show the world what a game could look like in the future once raytracing is added to the core capabilities of a PC, or maybe even a console one day. We teamed up with Nvidia and our friends at LucasFilm, the ILM X-Lab, to make a short film that demonstrates the core capabilities of raytracing in Unreal Engine. It's an experimental piece, but it shows the kind of features we'll add to the engine over the next year or so.

We've added support for what we call textured area lights, which is the same way we would light movies. You can see multiple reflections. You can see on the character, when she's carrying her gun, the reflection of the back of the gun in her chest plate. It's running on an Nvidia DGX-1, which is a four-GPU graphics computer they make. But as you know, hardware gets better every year. Hopefully one day there's a machine that can do this for gamers as well as high-end professionals. It's beginning to blur the line between what a movie looks like and what a game can look like. We think there's an exciting time ahead.

One thing we've been interested in over the years is digital humans. Two years ago we showed Senua, the Hellblade character. To this day, that's pretty much state of the art. But we wanted to see if we could get closer to crossing the uncanny valley. She was great, but you could see that the facial animation wasn't quite there. The details in the skin and the hair—it was still a fair way from crossing the uncanny valley.

Video is available on YouTube: Siren, alone (42s) and Siren Behind The Scenes (52s), and Creating Believable Characters in Unreal Engine (56m31s).

Related: Microsoft Announces Directx 12 Raytracing API

Original Submission

MrPlow writes:

Submitted via IRC for SoyCow3941

"An essential part of the clinical imaging pipeline is image reconstruction, which transforms the raw data coming off the scanner into images for radiologists to evaluate," says Bo Zhu, PhD, a research fellow in the MGH Martinos Center and first author of the Nature paper. "The conventional approach to image reconstruction uses a chain of handcrafted signal processing modules that require expert manual parameter tuning and often are unable to handle imperfections of the raw data, such as noise. We introduce a new paradigm in which the correct image reconstruction algorithm is automatically determined by deep learning artificial intelligence.

"With AUTOMAP, we've taught imaging systems to 'see' the way humans learn to see after birth, not through directly programming the brain but by promoting neural connections to adapt organically through repeated training on real-world examples," Zhu explains. "This approach allows our imaging systems to automatically find the best computational strategies to produce clear, accurate images in a wide variety of imaging scenarios."

Source: New artificial intelligence technique dramatically improves the quality of medical imaging

Original Submission

cocaine overdose writes:

From the New York Times:

The [Bronx] court sealed the case file, hiding from view a problem so old and persistent that the criminal justice system sometimes responds with little more than a shrug: false testimony by the police.

[...] "Behind closed doors, we call it testilying," a New York City police officer, Pedro Serrano, said in a recent interview, echoing a word that officers coined at least 25 years ago. "You take the truth and stretch it out a little bit."

[...] An investigation by The New York Times has found that on more than 25 occasions since January 2015, judges or prosecutors determined that a key aspect of a New York City police officer's testimony was probably untrue. The Times identified these cases — many of which are sealed — through interviews with lawyers, police officers and current and former judges.

In these cases, officers have lied about the whereabouts of guns, putting them in suspects' hands or waistbands when they were actually hidden out of sight. They have barged into apartments and conducted searches, only to testify otherwise later. Under oath, they have given firsthand accounts of crimes or arrests that they did not in fact witness. They have falsely claimed to have watched drug deals happen, only to later recant or be shown to have lied.

[...] Many police officials and experts express optimism that the prevalence of cameras will reduce police lying. As officers begin to accept that digital evidence of an encounter will emerge, lying will be perceived as too risky — or so the thinking goes. [...]

Yet interviews with officers suggest the prevalence of cameras alone won't end police lying. That's because even with cameras present, some officers still figure — with good reason — that a lie is unlikely to be exposed. Because plea deals are a typical outcome [...]

"There's no fear of being caught," said one Brooklyn officer who has been on the force for roughly a decade. "You're not going to go to trial and nobody is going to be cross-examined."

[...] Police lying raises the likelihood that the innocent end up in jail — and that as juries and judges come to regard the police as less credible, or as cases are dismissed when the lies are discovered, the guilty will go free. Police falsehoods also impede judges' efforts to enforce constitutional limits on police searches and seizures.

Original Submission