This AI Can Diagnose a Rare Eye Condition as Well as a Human Doctor

Diagnosing medical conditions is among the more classic examples of actually useful, achievable real-world machine learning. Machines have data, lots of it, and they have the capacity to process all of that data in ways that humans can't. Crucially, machines should be able to pick up on-the-edge cases, the rarest diseases that may go undiagnosed for simple lack of experience on the part of even the most exceptional doctors. Here, machines are to augment humans, rather than replace them.

To this end, a group of Chinese ophthalmologists and computer scientists has demonstrated a machine learning algorithm for identifying congenital cataracts, a rare eye disease that's nonetheless responsible for some 10 percent of all vision loss in children worldwide. The algorithm was able to catch the disease with accuracies exceeding 90 percent, putting it on par with individual human ophthalmologists. The new algorithm is described in the current issue of Nature Biomedical Engineering.

It's oft remarked that medicine is an art as much as it's a science. The interaction between a doctor and a patient is difficult to quantify—it's complex and shaded by intuition. Of particular concern when it comes to applying computational methods to medicine is the prospect of being wrong, which can have a far different meaning when that wrongness is the result of bad calculations rather than human judgement.

A doctor that makes a mistake is a human—possibly a grossly negligent human, but not necessarily—while a machine that makes a mistake is itself a mistake. A bug.

But the case for a machine role is there. "Machines have the advantages of automation, objectivity and precision, but the human ability to communicate and interact effectively is indispensable for medical treatment," study co-author Haotian Lin, a professor of ophthalmology at Sun Yat-sen University, told me.

"For doctors, technology is not sufficient to determine the best course of treatment with 100 percent certainty, and doctors should therefore make good use of the machine’s suggestion to identify and prevent the potential misclassification and complement their own judgment," he said. "The results of our comparative analysis showed that both artificial intelligence and human intelligence have strengths and limitations."

Missed or mistaken diagnosis is thus common among rare-disease patients, and this is especially true among large populations in developing countries, like China. Congenital cataracts are an especially compelling test-case because of the possibility of reversing the illness given timely intervention and rigorous follow-up care.

The algorithm is based on convolutional neural networks (CNNs), a class of machine learning models that attempts to imitate the neural processing that occurs in the visual cortex of animals. CNNs are widely used for visual recognition tasks but also other domains, like playing Go, natural language processing, and drug discovery. The very basic idea is of feeding the network large sets of images, including those of known congenital cataract cases, until eventually it learns an abstract representation that can be used to successfully analyze new images.

Here the researchers came up with three different networks useful for three variations on the cataract recognition task. The first is used for screening patients from the general population; the second is used for "risk stratification" among cataract patients; the third is used in assisting ophthalmologists with treatment decisions. All three are bundled up into a cloud-based platform known as CC-Cruiser.

The real-world test of the platform was based on 50 cases selected by an expert panel and consisting of a wide range of "challenging clinical situations." The performance of CC-Cruiser was stacked up against three categories of doctor: novice, competent, and expert. The 90 percent statistic cited above becomes more impressive when the results are broken down further. In normal cases, CC-Cruiser actually had no missed diagnoses and no false positives. No category of human doctor had that kind of performance. It only started to falter when tasked with making decisions about follow-up care, where the network registered a relatively large number of false positives.

CC-Cruiser is promising but not quite yet ready for IRL use. "Currently, our agent has been used in three non-specialized collaborating hospitals to further validate the feasibility of real-world clinical implementation," Lin said. "However, due to the need to respect and protect human life, medical fields always hold conservative and cautious attitudes when facing cutting-edge technology. Further rigorous clinical trials were still needed before we put the AI into regular clinical practice."

from This AI Can Diagnose a Rare Eye Condition as Well as a Human Doctor

Mathematician Proposes Blocking Tsunamis with Sound Waves

Usama Kadri, a mathematician at Cardiff University in the UK, has published new calculations in the open-access journal Heliyon demonstrating the possibility of neutralizing tsunamis with underwater sound waves. While actually implementing such a scheme would be enormously expensive and an enormous technical challenge, there aren't a whole lot of other tsunami defenses that don't basically just reduce to "getting the fuck to high ground before the wave hits." So, it's pretty novel.

The sound waves in question are more properly known as acoustic gravity waves (AGWs): vast underwater waves that travel at the speed of sound and are generated naturally by earthquakes and other geological events. In a sense, Kadri is then proposing fighting fire with fire. AGWs form naturally with tsunamis and act as subsurface precursors to the main event, affecting disturbances to the water column all the way from the surface to the seabed. AGW detection has recently been proposed as a an early-warning mechanism for tsunamis and rogue waves.

"Besides acting as tsunami precursors, AGWs can exchange and share energy with surface ocean waves," Kadri explains. This exchange occurs in an interaction known as the resonant triad, which is probably easier to just visualize.

The catch with the above setup, where a single smaller wave (an AGW) is used to drain energy from a much larger wave, is that it results in the creation of a second AGW and the two AGWs just wind up swapping energy amongst themselves rather than draining off energy from the tsunami. Kadri's answer to this is to start with two AGWs. The two AGWs moving at a much faster speed and in the opposite direction of the tsunami quickly sap away energy from the larger wave and scoot it far away.

The tsunami doesn't just die, but the interaction leaves it significantly weaker. And "significantly weaker" could mean saving hundreds of thousands of lives and billions of dollars in property damage, as in the case of the 2004 Indian Ocean earthquake and tsunami (Kadri's example). Unfortunately, this is all easier calculated than done.

"The amount of energy required to generate AGWs, given a realistic scenario, is probably much higher than the AGW energy, whereas the associated amplitude reduction is probably far less efficient," Kadri concedes. "Thus, there is a need to improve the interaction efficiency further."

But there's a deeper problem. The wavelengths of the AGWs required to have an effect on the tsunami are so long that it's difficult to imagine them being produced mechanistically at all. Kadri imagines that it may be possible to harness the same AGWs produced by the earthquake that produced the tsunami to somehow reflect back at the tsunami in modulated form. That's a pretty vague possibility.

"While detection is relatively straight forward," Kadri concludes, "the mitigation of tsunamis requires the design of highly accurate AGW frequency transmitters or modulators, which is a rather challenging and ongoing engineering problem.

from Mathematician Proposes Blocking Tsunamis with Sound Waves

How Breaking One of the Most Basic Physical Laws Might Explain Dark Energy

A group of theoretical physicists from France and Mexico has offered a fun new what-if for dark energy, one of physics' most profound outstanding mysteries. As described in the current Physical Review Letters, this proposed solution involves violating a key principle in our most basic understanding of fundamental physics: the conservation of energy. In this new framework, dark energy just represents the sum total of many tiny leaks of non-conserved energy spread throughout the universe. It's a bit weird.

To recap, the law of conservation of energy states that energy in an isolated system can neither be created nor destroyed. It can only change forms. Chemical energy turns to electrical energy in a battery; kinetic energy turns to thermal energy via friction; potential energy turns to kinetic energy as the skydiver leaves the plane. Both old-school Newtonian physics and relatively new-school relativity depend on this law.

Meanwhile, dark energy is the vast something in the universe that adds up to about 68 percent of all energy in existence. Crucially, dark energy is a repulsive force—it pushes things apart. Because of dark energy, the universe is not only expanding, but is accelerating in that expansion. Space itself is being thrown apart, and the more "apart" it becomes, the faster the expansion occurs. Empty space begets dark energy begets empty space. Eventually, that's all there will be: cold, empty space.

Our awareness of dark energy is only about two decades old. In the 1990s, observations made with the Hubble Space Telescope tipped astronomers off that the universe seems to be accelerating in its expansion. The repulsive force responsible for this was dubbed dark energy. Incredibly, circa 1917, Einstein had battled with an explanation for dark energy without ever having been aware of its IRL existence. At the time, he just needed some force that would counteract gravity at cosmic scales to allow the universe to remain in a more or less static state and not immediately collapse in a big crunch.

This was Einstein's cosmological constant. In short time, new astronomical observations of cosmic expansion would allow him to abandon the idea, and so it remained tucked away until the 1990s discovery of dark energy. The accelerating universe and its implied dark energy happened to align with Einstein's shelved cosmological constant and so the idea was reborn. Einstein had imagined a repulsive force permeating empty space, and we now know that this force exists.

This repulsive force has a tempting explanation. We know that empty space breeds a curious fizz of "virtual particles" just because quantum physics forbids actual empty space. So, as the universe expands and more would-be empty space is created, the more repulsive energy exists. That's nice, but as it turns out, the predicted repulsive energy of these virtual particles is vastly smaller than what's required to explain the observed expansion of the universe. Mystery not-solved.

Finally, we return to the current paper. "Ever since the discovery of the acceleration in the Universe’s expansion, almost two decades ago, there has been a puzzlement about the strange value of the corresponding cosmological constant Λ , the simplest, and so far most successful, theoretical model that could account for the observed behavior," the authors note. "The origin of this puzzle is that, within the usual framework, the only seemingly natural values that Λ could take are either zero or a value which is 120 orders of magnitude larger than the one indicated by observations."

In the framework proposed by the new paper the cosmological constant isn't so constant; Λ changes in accordance with the aforementioned tiny energy leaks as "a record of the energy-momentum nonconservation during the history of the Universe." It's when the universe reaches current scales that regular matter becomes so diluted by dark energy that Λ starts to look constant.

How these energy leaks are supposed to happen is more difficult to explain. The basic idea is to first imagine gravity as a continuous sheet, a field that has a value at every point all the way down to the infinitely small. Quantum physics is concerned with what happens when we take seemingly continuous fields like this and make them instead "grainy" or quantized—for example, light waves becoming the individualized packets of light known as photons. We don't really have a quantum theory of gravity yet, but if we can imagine at some point that continuous gravitational fields become individual particles, we can imagine that some energy may be lost in that translation.

So, if this energy is loss is real—and that's a big ol' if—we can come up with a pretty convenient place for it to go that helps explain dark energy. Unfortunately, this whole idea just throws us up against another mystery, which is quantum gravity itself. Proving or disproving dark matter-as-energy nonconservation is a long ways off.

from How Breaking One of the Most Basic Physical Laws Might Explain Dark Energy

Laser Squeezing Pushes Tiny Metal Plate to New Temperature Low

A couple of weeks I ago I wound up with frostbitten fingers after an ill-advised cold-snap bike ride. It was minor in the grand scheme of frostbite, but I wound up with almost-alarming blue tinge and that characteristic thaw-pain that feels like all of your cells are about to burst open. Taking the actual wind chill together with the apparent wind chill of the moving bike, I'd say we were down to about 10 degrees Fahrenheit.

That's nothing, of course. The average temperature of the universe is around −454.76 degrees Fahrenheit, while the current world record for laboratory-based coldness is about 0.0000000001 of a kelvin (0 kelvins being absolute zero). That record doesn't mean very much in terms of my fingertips, however—it was achieved by cooling the nuclear spins of a piece of rhodium metal. So: subatomic cold.

Macroscale cold, or cold that we can see with our own eyes, is a different matter. At near-zero scales, it's much more difficult to achieve. Rather than dealing with the nucleus of a single atom, we're cooling many whole atoms together, which means restricting the innate motions—read: quantum fluctuations—of many atoms together. To this end, physicists have found a new way of cooling macroscale objects to below previously established limits via a technique known as "squeezed light," according to research published recently in Nature. It is cold that should not be.

The macroscopic object in question was an aluminum plate about 20 micrometers in diameter, or a bit less than half the width of a human chair. This isn't exactly fingertip-scale, but it's an entirely different realm compared to subatomic particles.

Image: Teufel/NIST

The challenge in cooling an object this large lies in what seems to be a fundamental barrier: that quantum backaction limit. This limit is a consequence of the uncertainty found in quantum systems, which are characterized by random fluctuations rather than the well-established positions and velocities we're used to in our everyday not-so-quantum world. The particles that make up our metal plate are always fidgeting, and this motion—this noise—is in defiance of true coldness, and, ultimately, absolute zero, where all motion ceases at every scale. Noise is heat.

The physicists behind the current paper were able to cool beyond the apparent limit of particle noise thanks to lasers. It seems counterintuitive—a laser should add energy to a system, and so it should add heat to the system.

The basic idea is to start with a microwave cavity, which is a small space within which light waves tune themselves to match the natural resonant frequency of the space itself. This cavity contains the aluminum plate. Apply a frequency of light to the microwave cavity that's below its natural resonance and higher frequency light particles that match the resonant frequency of the cavity start appearing. As the space fills with microwaves, these particles leak out. And every time a photon leaks away, it steals a bit of mechanical energy from the plate, cooling it.

This cavity setup has been used before, but what the current research adds is the aforementioned "squeezing." Like all waves, light waves have a property called phase, which is how the waves are scooted forward and backward in time. Phase is subject to quantum fluctuations just like any other measurable property of particles, but it's possible to create waves that are super-regular with respect to phase, and this offers a way of transferring the quantum fluctuations from other properties of the light particles to the fluctuation-stripped phase property. This bit of cheating is how the researchers got beyond the backaction limit.

“We are squeezing the light at a ‘magic’ level—in a very specific direction and amount—to make perfectly correlated photons with more stable intensity," offers NIST scientist John Teufel, who led the experiment, in a statement. "These photons are both fragile and powerful.”

This isn't trivial research. Being able to cool larger and larger systems means being able to explore how quantum physics behaves at larger and larger scales, which is key to tasks like quantum information processing.

from Laser Squeezing Pushes Tiny Metal Plate to New Temperature Low

Spending Time Around Traffic Is Literally Destroying Your Brain

The highways around where I live in Portland, Oregon are like congestion voids—black holes or boundless oceans. You go in hoping to reappear elsewhere in the city or beyond, but without any assurance of exiting at all; time and space cease to have any meaning as waves and waves of Subarus roll toward the horizon. As once-plentiful provisions of Planet Money podcasts dwindle, you can feel madness setting in.

As it turns out, being around traffic for extended periods of time may have a very real effect on the human brain. This is according to Health Canada-funded research published this month in the Lancet that looked at the neurological health of two large-scale populations living in Ontario consisting of several million adults each. The study found that those individuals living closest to busy highways suffered from significantly increased rates of dementia, a symptom of irreversible neurodegeneration.

Specifically, the study found that up to one in 10 cases of dementia could be attributed to traffic exposure. This backs up earlier research finding that living near roadways—and the associated air pollution—can be tied to "insidious effects on structural brain aging."

The Health Canada study looked at all adults aged between 20 and 85 living in Ontario, a population of about 6.6 million people. It used postal codes to determine proximity to roadways and health records to determine incidence of dementia, Parkinson's disease, or multiple sclerosis. No correlation was found between the latter two afflictions and living around traffic.

Dementia risk, however, did vary with proximity to busy roadways. Those living within 50 meters of a busy road were about 7 percent more likely to develop dementia. At 50 to 100 meters, the increase went down to about 4 percent, while 100 to 200 meters led to a 2 percent increase. At greater distances, no significant increase was found.

By controlling for two common air pollutants—nitrogen dioxide and fine particulate matter—the researchers were able to negate some but not all of the increased risk. This points to the likelihood that the increased dementia risk is due to a combination of factors, possibly including the increased noise levels found around traffic.

The study controlled for complicating factors like socioeconomic status, education levels, BMI, and smoking, but as an observational study—where the variable of interest isn't under the direct control of the researchers—it can't make strict claims of causality. Still, given the above controls, it's hard to imagine what else the culprit could be.

So, what we're facing is a major public health concern. In a separate Lancet commentary, Lilian Calderón-Garcidueñas, a researcher at the University of Montana studying the neurological effects of air pollution, offers this conclusion: "the robust observation of dementia involving predominantly urban versus rural residents, opens up a crucial global health concern for millions of people... The health repercussions of living close to heavy traffic vary considerably among exposed populations, given that traffic includes exposures to complex mixtures of environmental insults..."

"We must implement preventive measures now, rather than take reactive actions decades from now."

from Spending Time Around Traffic Is Literally Destroying Your Brain

Ecologists Offer New Explanation for Mysterious Namibian Fairy Circles

As do most things that are highly symmetrical and otherwise exacting, the so-called fairy circles populating the grasslands of Namibia look intentional. They look careful and, well, supernatural. They just aren't messy enough to be regular-natural.

It helps of course that the Namibian fairy circles don't have a readymade natural explanation. A fairy circle consists of a region of grassland that's completely devoid of grass and bordered by a bushy circumference of unusually robust grass growth. Fairy circles have a lifecycle, becoming noticeable at about two meters in diameter and growing to be up to 12 meters in diameter before eventually giving in to the surrounding grasslands and assimilating. This can take up to 60 years.

Researchers at Princeton University, led by ecologist Corina Tarnita, have come up with a new explanation for the circles based on model simulations. The circles, according to Tarnita and colleagues, exist thanks to a complex interplay between subterranean insect activity and the emergent self-organization of the plants themselves. The fairy circles are indeed the work of nature, but it's a nature based not so much on messes, but on how seeming messes at small scales can yield unexpected—supernatural even—organization at larger scales, a phenomenon known generally as emergence.

The group's work is described in the current issue of Nature.

The fairy circles are the center of a long-standing debate. On the one hand, some ecologists attribute the circles to the work of "subterranean engineers"—termites or ants whose excavations lead to the disappearance of grass as vegetation is chewed up in the process of burrowing. Complex large-scale patterns can be found in termite mounds elsewhere, so we might imagine the subterranean version of a termite city to be similarly symmetrical. Moreover, according to a 2013 study published in Science, sand termites have been found in 80 to 100 percent of all fairy circles.

Image: Tyler Coverdale

But, on the other hand, some ecologists are skeptical of this explanation and point out that correlation does not imply causation. The presence of sand termites doesn't really prove anything, nor do termites explain the large size of the circles and their regular spacing across the grassland. To really explain things, we need to look to the grass itself. In particular, we need look at competition among different species of grass.

From this perspective, it seems likely that the circles represent nutrient reserves for the larger species of grass that populate their edges. The spacing among circles is then the result of competitive interactions between the tall grasses that depend on each circle. This hypothesis would seem to explain why the circles are found only within regions with just the right amount of rain, and, thus, just the right amount of nutrients.

Tarnita and colleagues basically argue that both perspectives are right. Simply, they were unable to generate the phenomenon in model simulations with either hypothesis alone. "These multi-scale patterns and other emergent properties, such as enhanced resistance to and recovery from drought, instead arise from dynamic interactions in our theoretical framework, which couples both mechanisms," the group writes.

So, in conditions where nutrients are scarce, we can imagine patches of grass dying via intergrass competition, leaving circles that get progressively larger until the nutrient deficit is overcome. The circles then allow rainwater to collect in them, which supplies more nutrients, which benefits everybody, including the termites. It all works out.

So, the circles aren't footprints of the gods, as in the local lore of the Himba people, but are instead examples of natural symbiosis that just happens to have elegant results. "Supernatural" is in the eye of the beholder.

from Ecologists Offer New Explanation for Mysterious Namibian Fairy Circles

Siri Saves Another Life: Mountain Bike Pro Andrew Cho

Andrew Cho made his name doing things on a mountain bike that you really aren't supposed to do on a mountain bike—like shredding BMX tracks, launching off of roofs, and doing backflips. But the time he really got into trouble was while home alone following a dinner with friends. After feeling kind of dizzy throughout the evening, he stood up only to immediately collapse. He was paralyzed from the neck down.

According to a GoFundMe page set up to help with Cho's medical bills, the mountain biker, now a marketing manager for bike maker GT, managed to inch toward his phone using only his chin. He used his tongue to activate Siri, with which he was then able to call 911. Soon after, he was rushed into emergency surgery. Cho had broken a blood vessel in his C3 and C4 vertebrae, which had caused the paralysis.

By now, Siri's kind of getting a reputation for saving lives. Last spring, an Australian woman used Siri to call an ambulance after her baby had stopped breathing. In 2015, a man in Tennessee used it to call for help after being pinned underneath a pickup truck. In another incident, a toddler was able to call 911 using Siri after her mother cracked her head against a table after fainting. It would appear that this is a thing.

In a sense, it's been a thing for a long time. Medical alarm systems have been around since the 1970s courtesy of Wilhelm Hormann and his notion of hausnotruf, or "home alert." Hormann's theoretical concept was brought to the real-world market in 1975 by a company called the American International Telephone Company. Its Phone Care "emergency dialer" consisted of a wearable button that, when activated, communicated with a base station that would then dial a number on a rotary phone and deliver a pre-recorded message. The device went for $795 at the time.

Read More: Why Amazon's Alexa Ran Away With CES

Voice recognition together with portable devices has democratized hausnotruf. None of the scenarios mentioned above involved senior citizens, the target market of medical alarm systems. These are just relatively young and healthy people with phones, and relatively young and healthy people with phones get into tight spots, too.

There are signs that the markets are merging, however. While there's still an enormous marketplace for old-school medical alert systems featuring a button and a base station, apps are naturally stepping in. For as little as $6.99 a month, the the OnCall Defender Panic Alarm will equip your smartphone with a GPS-enabled panic button. Rather than dial 911, it feeds directly into a staffed command center with "a direct connection to local law enforcement." It's kind of like a security system that you wear around. The OnCall Defender is advertised as the first of its kind.

Smartphones come with something else that may be able to help in medical emergencies, even when the victim is unable to speak: an accelerometer. Already, apps exist, such as Emergency Fall Detector on Android, that can supposedly detect falls and contact help according to the user's predefined instructions.

Of course, the idea of our phones becoming nurses and effectively following us around scanning for trouble is its own kind of dystopia—probably one that an extreme mountain biker would otherwise abhor.

Get six of our favorite Motherboard stories every day by signing up for our newsletter.

from Siri Saves Another Life: Mountain Bike Pro Andrew Cho

A Search Engine for Programming Language Syntax Is a Pretty Good Idea

The current search engine for programming language syntax is Google. Knowing how to search for information is a key skill in knowing how to program at all. You can know all of the algorithms and a half-dozen programming languages inside-out, but you will nonetheless be searching for how to do something at some point, whether it's related to some brand-new or super-obscure functionality or to how to translate some feature or another in one programming language to another language.

In other words, knowing how to program has a lot to do with knowing how to access information—an acute awareness of how and when to learn.

This learning might occur in hyperdrive if you're the sort of programmer that's either obsessively learning new things just for the sake of it—which is a whole lot of programmers—and-or has to learn new things to apply them to a new project or task. For a recent project, for example, I needed to use a machine learning framework that's implemented in a kind of obscure language called Lua, which is like a super-lightweight version of Python. I watched a couple of videos, but mostly I was inferring syntax from other Lua code and Googling things like "Lua for loop break."

A computer engineering student at Queen's University in Ontario named Anthony Nguyen has released what he hopes will replace Google for the syntax searching needs of software developers. It's called SyntaxDB, which Nguyen hopes will, "one day become the world's fastest programming reference."

I did some cursory searching and the SyntaxDB interface is pretty nice. Searching, say, "Java iterator" gives a few results for Java control flow structures and a sidebar menu for further exploration within the Java language. Those results are all internal to SyntaxDB and lead to reference materials written exclusively for the database. It's more like a multi-language quick reference manual than a search engine.

As for those results, I didn't actually get a page for Java iterators, which seems bad and also seems like a limitation of relying on in-house content rather than trying to exploit the bottomless stockpiles of programming language documentation that already exist on the internet. Just writing your own reference materials seems easier in the short-term, but maybe not all that sustainable.

The big advantage of SyntaxDB that I can see is in the brevity of the documentation. An entry on for-loops isn't going to give you every detail, or most any details at all, just what one looks like and its basic usage. And that's often all you're looking for—what a known concept or feature looks like in an unfamiliar programming language.

SyntaxDB has an integration with the general internet search engine DuckDuckGo that seems more useful than using SyntaxDB directly. Here, you can enter a syntax search term in the normal search bar and DuckDuckGo will give you the quick and dirty SyntaxDB documentation, if it exists, along with the usual list of internet search results. If it doesn't exist, DuckDuckGo will find the next best thing.

Will I use it? Mayyyybe. I'm already pretty good at cultivating quick and dirty answers from complicated documentation, and, like a lot of people that write code, I also often enough turn to the the vast stockpile of ad hoc documentation found in the programmer Q&A forums of StackOverflow. That's something Google can give me that I'm not sure could ever be replicated by SyntaxDB.

from A Search Engine for Programming Language Syntax Is a Pretty Good Idea

Fish Are Having a Real Hard Time in Space

Fish traveling aboard the International Space Station in 2014 experienced a near-immediate reduction in bone density upon encountering the microgravity environment of orbit. This is according to research published recently in Scientific Reports by a team of biologists at Tokyo Institute of Technology who conducted remote imaging experiments on newly-hatched medaka fish launched into space.

The general findings are concerning but not all that surprising. The dramatic effects of microgravity on bone density have been observed in human astronauts aboard the ISS, where bone deterioration begins after about 20 days in orbit in a process resembling the sort of osteoporosis more often associated with old age. The mechanisms behind this, however, are still being explored, both for the sake of long-term space travel and for treating osteoporosis here on Earth's surface. And so we have medaka fish, whose process of skeletogenesis is similar to our own.

"Under microgravity, there are several changes in the animal body, such as fluid shift, increase in blood pressure, and dizziness," Akira Kudo, the study's lead author, and colleagues explain. "In particular, bone mineral density is decreased under microgravity; but it is unclear how osteoblasts or osteoclasts respond early in orbit."

To better understand these biological effects of “microgravitational stress,” there are two varieties of cell that need to be observed: osteoclasts and osteoblasts. The prior are responsible for breaking down bone tissue, a key role in repairing and maintaining bones, while the latter secrete the matrix used in bone formation.

The researchers were able to observe the activity of these cells live from a remote lab at Tsukuba Space Center using fluorescence microscopy. Basically, they created transgenic fish that would glow under different wavelengths of light. The fish were actually hatched at Baikonur Cosmodrome, the Soyuz rocket launch site in Kazakhstan, after being transported from Japan as eggs by Kudo and co.

The fish spent the first six weeks of their strange lives at the launch site before being embedded in a special gel for their voyage aboard Soyuz flight TMA-10M. They then spent the next two months being reared aboard the ISS. The first eight days of that stay were spent underneath a fluorescence microscope as researchers on Earth observed the fish's bone cells misbehaving in real-time. These observations were compared to an Earth-bound control group of medaka fish.

Gene expression markers for both varieties of bone cell increased significantly from the control group of medaka on Earth. On the very first day, this increase was seen in osteoclasts, where it persisted for up to eight days. In osteoblasts, the increase came four days following arrival at the ISS. The osteoclast increase was seen in two bone-specific genes: osterix and osteocalcin. Normally, these appear in fish development at different phases, with osterix appearing several days before osteocalcin. Here, they appeared simultaneously, offering a new clue into the fundamental mechanism behind spaceflight-related bone loss, and, perhaps, the widespread but poorly treated affliction of old-age, osteoporosis.

Of course, we're still just talking about lab-bred fish on a space station, so obviously more research is needed if we're to ever apply this knowledge to humans. But tucked into the paper, Kudo and his team tease at the possibility of their research opening up a whole new scientific field: "gravitational biology."

from Fish Are Having a Real Hard Time in Space

Hack This: Make a Photo Filter with Machine Learning

For a while, people really loved complaining about Instagram filters. They make photos look different, but always in the same way. They encourage lazy photography. A cheap stylistic substitute for thoughtful composition. Etc. It was another whole big hand-wringing session about authenticity. Do people still tag things with "#nofilter"?

But to call image analogizing "filtering" is to sell the associated framework short. In the words of its New York University-based creators, it's "processing images by example." "Rather than attempting to program individual filters by hand, we attempt to automatically learn filters from training data," the project's website explains.

The idea: Give the Image Analogies framework three images and it will teach itself what makes the first two images similar, and then apply that similarity to the third image as a filter. The results are unpredictable and frequently very cool. Fortunately, you don't have to be a machine learning whiz to do it.

That said, there is some set-up, which is the hardest part of making your own image analogy. I'll walk you through it below. We'll be using a Python implementation of the original NYU method written by programmer Adam Wentz, whose other projects include Huge Wall of Porn, gif hell, Oldstagramme, and more fun stuff.

0.0) Resources

You don't need a GPU and a ton of memory to use Image Analogies, but it helps. Most of my experiments were done with Amazon EC2 instances that come with mostly all of the needed math/machine learning software preinstalled, and GPUs to run it on. (GPUs and their parallel processing abilities are key to the sorts of computations involved in machine learning.) Running on a remote machine also has the advantage of not completely inundating your own computer's processors, which can happen fast.

All that said, I'm not going to explain the whole process of getting started with EC2 instances and interfacing with a remote shell via the command line. Maybe in another Hack This edition. Let's just assume for the sake of this tutorial that everything is being computed locally on the machine in front of you. That will work.

0.1) Software

Image Analogies will work with either of the two big machine learning libraries TensorFlow and Theano. If you have a GPU to work with, you'll probably want the latter, while, if not, you'll need to be using TensorFlow. Another machine learning library called Keras is then needed to run on top of either Theano or TensorFlow. Yeah, I know: This is already getting to be pretty messy. But hold tight for a sec.

Going forward, I'm going to assume that we're working with TensorFlow because there's actually quite a bit more to getting going with GPU support, mostly having to do with the installation of CUDA, which is the software/platform (yes, another one) that lets you use your Nvidia GPU for these kinds of computations in the first place.

As for getting going with TensorFlow, you can follow the official guide here. It's not too bad and can be accomplished using pip like any other Python package.

The Image Analogies framework itself can be installed using pip by the simple command: pip install neural-image-analogies. This installation should take care of all of the framework's dependencies, including Keras and TensorFlow. You might want to do this in a Python virtual environment, which will keep the Image Analogies installation from possibly breaking a dependency chain elsewhere on your system.

1.0) Weights

With everything installed and theoretically working correctly, we can get to some actual machine learning. First, we need to download VGG16, which is a 16-layer convolutional neural network that's about the state-of-the-art in image recognition. This is what Image Analogies will use to make sense of our two input images. You can download a reduced form of VGG16 here. Note that it will need to be in the same directory from which you're running the Image Analogies script from.

2.0) Experiment

We're basically ready to go. All that's left is picking some images. Remember, we're taking two images, comparing them, and then applying the results of that comparison to the third image. You can be pretty clever with this.

From Wentz's Image Analogies Github page:

Mostly, I've just been throwing images together and seeing what happens. You can wind up with some cool patterns, at least. There are also a million different options and parameters you can run this script with (see the Image Analogies Github), which let you do everything from isolating specific layers of the machine learning model to tweaking detail levels and image scales. You could burn through an afternoon pretty easily just taking the "throw stuff together" approach (as below).

Once Image Analogies is installed per the above instructions, it's launched with the following command: python make_image_analogy.py first-image second-image third-image filename-prefix-for-output. The filename prefix is what the script is going to stick onto the beginning of every filename that it saves to your computer. These images should be in whatever directory you're running the script from. If you wanted to save the output to a different directory, you'd just prepend that to the output filename prefix, like: /otherdirectory/outputprefix. By default, you'll get sample intermediate images saved to your computer as the algorithm chugs along. If you wind up getting sucky output images, it's easy enough just to cancel the script before it finishes.

For me running this on a fairly standard-issue MacBook, each run takes about a half-hour. Running on a GPU-optimized Amazon instance knocks that down to five or 10 minutes. If you were interested in taking Image Analogies to the cloud, I'd suggest looking at the Go Deeper Amazon system image, which comes with all of the Image Analogies dependencies prebuilt and ready-to-go. It also has some pretty user-friendly documentation to get you started if, say, you have no idea what I mean by "Amazon system image" or EC2 or even GPU computing. Go Deeper even offers a remote desktop, which is pretty handy if you're not used to interacting with a remote computer via ssh.

We're obviously just dipping a toe into something much, much bigger here. But, like Google's Deep Dream, it winds up being a good entry-point into the bigger thing that is visual processing and machine learning, generally. And it will be a good stepping off place for Hack This to go deeper too, at least sometimes.

Read more Hack This.

from Hack This: Make a Photo Filter with Machine Learning

Much of the World Is Still Subsidizing Gas Prices Like 'Climate Change LOL'

It's axiomatic that if you really want to keep people from buying something you just tax the shit out of it. Cigarettes would be one example where an acutely dangerous product underlying an vast public health emergency was progressively made expensive enough to make its consumption financially painful if not prohibitive.

Climate change is an emergency, too, albeit one that's felt rather more collectively. It's often remarked that the only way we're really going to get control over it is through taxation, essentially. Carbon taxes. Make climate change personally expensive.

One example of this is directly taxing gasoline, something that's common across much of the Western world, including the United States. In many nations, gas taxes have been rising, but, according to a paper published today in Nature Energy, many other governments subsidize gasoline consumption instead. In fact, the gas tax global mean fell 13.3 percent between 2003 and 2015 as gas consumption has shifted toward countries that maintain gasoline subsidies or that have very low taxes.

The current paper, which comes courtesy of political scientist Michael Ross and colleagues at the University of California Los Angeles, is concerned with a fundamental problem that extends beyond gas taxes themselves. This is the inherent murkiness of assessing and verifying energy taxes, generally.

"Self-reporting by governments is often incomplete and unreliable," Ross writes. "Many taxes and subsidies are indirect, or hidden in the budgets of state-owned enterprises; moreover, the real value of taxes and subsidies changes over time due to inflation and currency fluctuations. Some countries announce reforms but either fail to enact them or nullify their impact with countervailing policies, as in the case of Brazil. Others try to remove gasoline subsidies quietly to avoid dissent."

So, Ross and colleagues took a simplified approach to quantifying energy taxes and subsidies by looking at gas-pump prices in different countries—a highly accessible metric—and comparing those to a global benchmark price. The result is a picture of implicit gas taxes/subsidies. "The price gap method allows us to measure what the IMF calls ‘pre-tax subsidies’, which represent the difference between the retail price and the international supply cost," Ross and co. explain.

The results can be seen below, where the red line represents the global benchmark price for gasoline. Lines above indicate taxation, those below indicate subsidization.

Image: Ross et al

The highest gas taxes are mostly found in Europe and North America, while the highest subsidies are found in oil-rich countries located in the Middle East and northern Africa. Countries with subsidies are most often those that depend economically on oil exports. Redistribution of state wealth, essentially

The current paper covers only pre-tax subsidies, which is a limitation of looking only at gas pump prices. Post-tax subsidies, which account for the high costs of environmental harm absorbed by a nation on behalf of the fossil fuel industry (undercharging for global warming, that is), are another matter. A paper published recently in the journal World Development came up with global post-tax subsidy total in 2015 of over $5 trillion, or nearly 6.5 percent of global GDP.

So, what now? Taxation is a powerful and cost-effective tool for limiting greenhouse gases, but it only works if half the world isn't actively working against it. Politics circa 2017 aren't exactly primed for boosting gas taxes anywhere, but we at least have a metric now to prevent nations from subsidizing gasoline consumption in secret.

from Much of the World Is Still Subsidizing Gas Prices Like 'Climate Change LOL'

Big Data's Unexplored Frontier: Recorded Music

While still a vast field, a huge part of machine learning exists for what may seem to be a relatively narrow subset of problems. These are problems involving visual processing: character recognition, facial recognition, the generation of trippy images dominated by populations of dogslugs, birdlegs, and spidereyes.

This isn't accidental. Image data is unique in its suitability for machine learning tasks. It naturally occurs as multidimensional arrays—tensors, really—of pixel data. It's more at the fringes of machine learning that audio data gets a turn. Part of the problem is that, despite the vast amounts of digital audio data that exists in the world, there is a relative lack of openly accessible computational datasets. There's pretty much just one, actually: the Million Song Dataset, which offers some 280 GB of feature data extracted from 1 million audio tracks. Musicology remains largely old-school.

A group of computer scientists based at University College London and Queen Mary University wants to change this. To that end, they've developed the Digital Music Lab (DML) system, a large-scale open-source framework for the analysis of musical data that, crucially, allows for the remote analysis of copyrighted materials (read: most music). "Our first installation has access to a collection of over 1.2 million audio recordings from multiple sources across a wide range of musical cultures and styles, of which over 250,000 have been analysed so far, producing over 3TB of features and aggregated data," they write in ACM Journal on Computing and Cultural Heritage.

The project started in earnest in 2014 with an initial workshop attended by 48 musicologists of varying technical backgrounds. It was basically a big discussion of the problem laid out above (the lack of computational resources for analyzing music data) and what might be done to solve it. The first part of that was just the dataset itself, which should contain basic metadata about individual samples and information about where they can be accessed. The second is the actual analysis.

This is the hard part. Each individual recording contains millions of data points that need to be analyzed for musical content, with the result being that each one of those recordings requires about as much processor time as the length of the recording.

"The data are richly structured: there is not just one analysis needed but several different ones, for example, for melody, harmony, rhythm, timbre, similarity structure, and so on," the researchers explain. "These create a rich set of relations within and between works and collections that is more complex than, for instance, typical textual data."

You're much better off just playing with the thing than having me paraphrase a technical paper about DML. It has a handy web interface here.

Or watch the demo:

Of course, musical recordings are fundamentally different than images. The vast ImageNet image recognition dataset doesn't consist of millions of careful compositions (or "compositions" at all), but instead of all manner of images gathered from the internet. Image recognition is ultimately a means for computers to better interface with the real-world. So, no, it's not quite a fair comparison, but the need for better musical analysis tools in the era of big data seems clear enough.

from Big Data's Unexplored Frontier: Recorded Music

Soon WiFi Will Be Able To Tell Not Just Where You Are, But If You're Breathing

WiFi networks soon may be able to tell not just where you are, but whether you're even breathing. This is according to research published this month in Computer by a trio of researchers at Peking University that describes a sensing method based not on the conventional received signal strength (RSS) model—which provides information about an environment based on how that environment attenuates a signal—but on an alternative known as channel-state information (CSI), which provides a much richer picture of electromagnetic waves as they bounce around an indoor space.

RSS positioning has been around for about 15 years and was developed into a project called RADAR by researchers at Microsoft. Its operation is pretty crude. Take an indoor space, and then map it out according to WiFi signal strength at different locations. Stick all of this data into a table, and when it comes time to locate an actual device in that space, it's just a matter of matching the observed signal strength to the corresponding location in the table. It's cheap, at least.

Using channel-state information for indoor sensing is already being actively explored, but what the Peking researchers wanted to know is exactly what kind of precision it's capable of offering. To find this out, they applied what's known as the Fresnel zone model, which is easiest to just visualize:

Here's how the paper explains it: "Fresnel zones refer to the series of concentric ellipsoids of alternating strength that are caused by a light or radio wave following multiple paths as it propagates in free space, resulting in constructive and destructive interference as the different-length paths go in and out of phase." You can break any area of space into an infinite number of Fresnel zones.

Dan Wu and his colleagues at Peking University explain that any object that a radio wave encounters as it bounces around a space essentially splits it into two. One is reflected while the other travels on through the object (following a line-of-sight path). At the receiving end of the signal, the two paths recombine, leaving a superimposed signal. It's from the phase difference between the two signals that an intervening object can be inferred. Objects positioned in different Fresnel zones will reflect signals differently, resulting in interference patterns corresponding to different positions.

"We conducted indoor experiments with a pair of Wi-Fi transceivers and a metal cup to verify the Fresnel zones’ existence and to show that the received signal varies as expected when an object moves across the zones," Wu and co. write. A set radio frequency was chosen and the metal cup was moved at centimeter intervals in three directions within the Fresnel zone spanning the transceivers. The different positions resulted in the signals superimposed as expected.

So, we wind up with a sensing limit of (at least) a single centimeter. That should be precise enough to detect human respiration, but Wu and his team had to test it out. They found that if the subject was close enough to the line of sight path travelled by the signal between transceivers—and was thus interacting with the signal at its strongest—they could accurately detect breathing. If the subject was too far away from a transceiver or the line of sight path, not so much.

The researchers are nonetheless optimistic about their proof-of-concept: "In the shorter term, we envision the proposed theory accelerating the nonintrusive human-sensing field, enabling a wide spectrum of new applications in homes, offices, hospitals, warehouses, and more. In the longer term, we believe that synergizing communication and sensing capabilities in computing devices will fuel a revolution in both Internet of Things (IoT) and context-aware computing."

Tracking: It's barely even getting started.

from Soon WiFi Will Be Able To Tell Not Just Where You Are, But If You're Breathing

Hack This: Where To Write Code

The code itself, the symbols and strings that eventually dictate the behavior of a machine, is one thing, but properly preparing it and executing it requires a dedicated venue. Your operating system's standard text editor—TextEdit on Mac and Notepad on Windows—will really only get you so far before becoming a limiting factor. These stock tools are generally meant for writing text intended for human consumption and offer few of the even most basic coding features, including looking the part.

Unfortunately, the world of code editors is a complete zoo. Some editors are elaborate Photoshop-like behemoths offering platforms suitable for building vast software projects, while others are esoteric and minimal, rewarding only those with patience and willingness to cope with a steep learning curve. Some editors are suited for specific programming languages and platforms, like Swift and iOS, while others aim to be general but extensible. A few are completely meaningless beyond one specific language, while a small set of code editors become programming languages unto themselves. So, given all of the noise, where should one even start?

I think I can help. What follows is a field guide intended for those most likely to be asking the above question: the code curious, beginners, advancing novices.

1.0) Supercharged text editors: Sublime Text and Atom

For the type of coding described in the Hack This series, this is the ideal category of text editor. It's dominated by Sublime Text and Atom. The latter is free, while the prior theoretically costs money but can be evaluated indefinitely as nagware (you should pay for it!). When you first open one of these editors, at first glance it may just seem like an iteration of TextEdit/Notepad with better colors. It's when you start digging around in the menus that the depth and capabilities become clear.

For one thing, both come equipped out of the box with syntax highlighting for pretty much any programming language you're likely to be interested in. This might seem kind of trivial when you're starting out, but once you know a language a bit and what a script or program in that language should look like, highlighting becomes immensely valuable in making code readable. Eventually, monochrome code will just look wrong, like English written without paragraph breaks or punctuation.

Editors in this category are extensible via package systems. The package ecosystem for Sublime Text is vast, with installable packages offering new editor capabilities ranging from those enabling regular word processing (like word count and spell check), HTML templating engines (like Handlebars), Git integration, code quality tools, and well beyond. Whatever sort of coding you're into, it's possible to essentially build up your own ideal code editor via these third-party packages. Personally, I don't take enough advantage.

2.0) IDEs

In my early-days community college programming courses, I starting out writing code within the beast that is Visual Studio, Microsoft's integrated development environment (IDE) geared toward developing full-on software for the .NET framework. Within Windows, this is the natural home for writing old-school PC software in C++, C, and C#. In school, this meant making things like prime number checkers, record collection organizers, and Colossal Cave Adventure clones.

IDEs are enormously powerful, and, unlike the aforementioned text editors, this power is flung at users out of the box. It can be intimidating. Most offer extensions for new programming languages and frameworks that essentially amount to new IDEs within IDEs. The Scala IDE, for example, is really a specialized version of the Java-centric Eclipse IDE (Scala is an extended version of Java). Similarly, Android Studio is really an extension of the more general Java IDE IntelliJ IDEA.

What drives me from a text editor like Sublime Text to a full-fledged IDE is usually organization and-or debugging. Sublime and Atom both offer the ability to manage multiple files and directories, but this just feels more natural in an IDE, particularly when things get really messy in contemporary web development, which demands extreme modularity and the inclusion of often many, many outside libraries.

When dealing with long, complicated programs, debugging becomes a far more involved process than just staring at a dozen lines of code for a few minutes until the error is magically revealed. Instead, it becomes neccessary to watch the code in action as it executes. This is what step-through debugging achieves. A debugger allows programmers to manually step through their code line by line as it executes, providing a means to observe side effects and state changes as they happen.

As such, bugs can be traced down to the lines where they occur, which is pretty neccessary when dealing with hundreds or tens of thousands of lines of code spread across many individual files. An IDE can allow programmers to even observe code execution at extreme low levels, including physical processor registers and memory addresses. Part of this low-level view is what's known as code profiling, where precise performance statistics can be collected about a given section of code. Bugs are, after all, more than just program-sinking errors, but include system-draining inefficiencies. You could say these are algorithm-level bugs.

As far as IDEs go, you'll find many arguments for the JetBrains family of software, including PyCharm for Python (above), WebStorm for JavaScript and website development, CLion for C and C++, and others. These are all highly worthy (but not free).

2.1) Xcode

Xcode is sort of its own class of IDE. This is where Apple pushes iOS development. It's built for writing code in Swift and its predecessor language Objective-C, the languages of iOS, and comes with specialized tools for simulating and profiling Apple devices. Part of its appeal is an emphasis on visual coding. As far as IDEs go, it's unusual in this friendliness to non-programmers. You can get pretty far in building an iOS app just through dragging and dropping. There's no Windows version.

3.0) Hardcore text editors: Vim and Emacs

Vim is hard enough to describe, let alone use. On its face, it's a hyperminimal text editor, a realm where even the mouse has no meaning and drop-down menus have yet to be invented. Menu actions instead are accomplished via text commands. Saving? That would be :w. Quitting without saving? :q!. Copy and pasting?

Glad you asked. Behold, the official Vim documentation:

*04.6* Copying text
To copy text from one place to another, you could delete it, use "u" to undo the deletion and then "p" to put it somewhere else. There is an easier way: yanking. The "y" operator copies text into a register. Then a "p" command can be used to put it.

Yanking is just a Vim name for copying. The "c" letter was already used for the change operator, and "y" was still available. Calling this operator "yank" made it easier to remember to use the "y" key.

If that makes sense, then congrats, you understand Vim. Based on years of reading Vim-related answers on Stack Overflow, this understanding is a point of pride among Vim users, which is putting it mildly. I kinda-sorta understand Vim.

Vim (and Emacs, its traditional competitor) offer more than esotericism for its own sake. There's immense power behind the maddening anti-interface (which is really a function of Vim and Emacs predating graphical operating systems and even PCs, generally). Interfacing with Vim is based entirely on keystrokes—even the cursor keys are shunned in favor of the h, j, k, and l keys—which winds up being really, really fast when you have the hang of it.

Consider that in a typical GUI environment most everything is accomplished by interrupting keyboard typing and interfacing with some menu element. Even if that element displayed front and center in a toolbar, the fingers are forced to leave and then return to the keyboard. In Vim, those same commands are achieved as a part of the same keyboard typing flow. That's the key word: flow.

Vim is extremely customizable and constitutes a programming language in itself. Much as we might enter a Python command into a Python interpreter or a Bash command into the Bash interpreter, we interact with Vim by entering Vim commands onto the Vim command line. As with Python and Bash, we can collect those commands into scripts, which are referred to in Vim configuration files.

There's a whole history here that I'll save for another post, but Vi (Vim's earlier incarnation) and Emacs are both now 40 years old and constitute the earliest text editors for any purpose, coding or otherwise. Both still come built into Unix-based operating systems, including OSX. Just type vim or emacs in the terminal.

So where should you write code? If you're asking the question, the answer is probably Atom or Sublime Text. Both are simple out of the box but will grow with you. Eventually, you'll be forced to pick something else up—like Rstudio for data analysis in the R language, or Xcode for app development—but when it comes to writing useful scripts for tasks like, say, webscraping or automating operating system tasks, you want a space that can offer features only when you need them and a space where you can focus on the code, not the environment.

Read more Hack This.

from Hack This: Where To Write Code

Nanogenerator Harvests Swipes To Power LCD Screens

There's a whole lot of energy out there that's just kind of hanging around. The brakes on cars and trains turn momentum into heat, for example, which we now have systems for recapturing and recycling. But there are many more examples of wasted "ambient" energy that we don't recapture. Even regular old walking around as bipedal animals is an inefficient process; the energy we expend in a single stride is greater than it would be given a perfectly efficient process.

Such is life, but nowadays we're surrounded by devices that don't require all that much power to operate. A couple of volts goes a long way. A newly developed nanogenerator, described this week in the journal Nano Energy, puts that into perspective, offering a means of converting the energy expended in a standard touchscreen swipe into sufficient power to light up a touchscreen.

The nanogenerator in question is what's known as a biocompatible ferroelectret nanogenerator, or FENG—a paper-thin sheet of layered materials including silver, polyimide, and a sort of giant charged molecule known as polypropylene ferroelectret. The layers of the FENG are loaded up with charged ions, which results in a construction that, when compressed, produces electrical energy.

The high-level picture is that the FENG winds up with really huge dipoles—magnetic poles of opposite charge—existing on its different layers, which then change in relation to each other as the material is deformed under pressure. This change results in differences in electrical potential, which is what gives us useful electrical energy.

So, we hear about self-powered devices kind of a lot. What makes this one interesting is that it's a new kind of device. That is, a FENG is not piezoelectric (electricity via squishing crystals) or triboelectric (electricity via certain kinds of friction).

The paper describes some advantages: "their simple fabrication allows for encapsulated low-cost devices. In view of the environment, health, and safety, the fabrication of encapsulated FENG avoids the use of harmful elements (e.g. lead) or toxic materials (e.g. carbon nanotubes), making it more attractive for biocompatible and perhaps even implantable applications."

The device also has the neat property of becoming more powerful when folded. In a statement, lead investigator Nelson Sepulveda explains: "Each time you fold it you are increasing exponentially the amount of voltage you are creating. You can start with a large device, but when you fold it once, and again, and again, it's now much smaller and has more energy. Now it may be small enough to put in a specially made heel of your shoe so it creates power each time your heel strikes the ground."

Sepulveda and co.'s current task is in developing technology that would allow for the transmission of energy generated by said heel strike into devices like headsets.

from Nanogenerator Harvests Swipes To Power LCD Screens

'Pizzagate' Conspiracy Meme Reaches Its Natural Conclusion—With Shots Fired

On Sunday, a North Carolina man was detained outside of a Washington, DC, pizza restaurant and concert venue after allegedly entering the establishment with an assault rifle and firing one or more shots. So far, there have been no reports of injuries, but the restaurant, Comet Ping Pong, and several others nearby were put on lockdown for several hours, according to the Washington Post.

Police Chief Peter Newsham initially told the Post, “At this point we do not believe that it was terrorist related. And it’s unclear right now what the motive is.” A DC Police statement released late Sunday noted that the gunman had explained in a post-arrest interview that he was "self-investigating" what's come to be known as "pizzagate".

It didn't take an interview confession to figure that out, however. Since before last month's election, Comet Ping Pong has been at the center of a highly bleak fake-news garbage fire, in which right-wing conspiracy types have come to believe that the restaurant is at the center of a grand-scale pedophilia ring involving none other than Hillary Clinton's presidential campaign chair John Podesta.

The substance of the conspiracy is nuts even by the standards of right-wing conspiracy memes. It involves systems of tunnels and back rooms and coded symbols and language. It turns out that if you substitute in "child sex abuse" for "pizza" you can find a whole lot of evil in the world. Pizzagaters congregated for a time at Reddit—see archived thread here—but the topic was banned last month.

Among its followers, pizzagate includes none other than Michael Flynn, the retired general tapped to be President-elect Donald Trump's national security advisor.

Snopes went deep on the conspiracy theory last month, but, to give some perspective, much of it revolves around the frequent appearance of pizza in Podesta and Co's hacked emails and the existence of ... a handkerchief featuring a "pizza-related" map on it. From Washington City Paper:

Why, they wondered, did Podesta have a handkerchief with a "pizza-related" map on it? And why did Podesta get so many emails about eating pizza?

The answer to any reasonable person would be that Podesta eats pizza sometimes. Indeed, Alefantis says, "pizza's always a big thing in politics." ... To the alt right, though, "pizza" became a suspected code word for illegal sex trafficking. ... In one 2008 email released by Wikileaks, Alefantis thanked Podesta for attending a fundraiser at the restaurant but regretted not making him a pizza. That drew amateur theorists' attention to the restaurant's murals, which they declared "creepy," and the sliding doors in front of the restaurant's bathrooms, dubbed "hidden rooms."

The DC Police statement identified the suspect as 28 year old Edgar Maddison Welch of Salisbury, NC. He's been charged with assault with a dangerous weapon.

Get six of our favorite Motherboard stories every day by signing up for our newsletter.

from 'Pizzagate' Conspiracy Meme Reaches Its Natural Conclusion—With Shots Fired

Most People With Depression Go Without Treatment

The vast majority of people suffering from depression worldwide are not receiving even minimally adequate treatment, according to a study published this week in the British Journal of Psychiatry based on surveys of more than 50,000 people in 21 countries. While the situation improves significantly in developed high-income countries, still only one in five receive adequate treatment where it's most available. Among the poorest countries, this figure falls to one in 27.

The findings are based on analyses of data collected as part of the WHO World Mental Health Surveys, a collection of 23 community surveys from 10 low- or middle-income countries (Brazil, Bulgaria, Colombia, Iraq, Lebanon, Mexico, Nigeria, People's Republic of China, Peru, and Romania) and 11 high-income countries (Argentina, Belgium, France, Germany, Israel, Italy, Japan, Netherlands, Portugal, Spain, and the United States). In all, the researchers, led by King's College London psychiatry professor Graham Thornicroft, tracked 4,331 people with depression.

In more than half of the cases (56 percent), survey respondents were aware that they needed treatment of some kind. Seventy-one percent of that group sought treatment on at least one occasion, while 41 percent of the treatment-seeking group received minimally adequate treatment (defined as either regular talk therapy with any professional or pharmacological treatment coupled with regular doctors visits).

Read more: Researchers Found Where Depression Lives in the Brain

According to the the Global Burden of Disease 2010 Study—which is cited in the current paper—major depressive disorder (MDD) ranks as the second leading cause of years lived with disability in the world. According to that study, MDD is the 19th most common disease in the world, and rates as among the most severe, based on the proportion of people with depression that will have it for a relatively long period of time.

"There is an increasing awareness that MDD can be reliably diagnosed and treated in primary care settings using antidepressant medications and/or brief structured psychological therapies, but substantial barriers exist to this care being delivered," Thornicroft and colleagues write. "These include supply-side factors (for example, policies to invest resources, and consequent scarce mental health services, community, and human resources), as well as demand-side issues (for example, lack of awareness of MDD as a treatable illness, and stigma and social exclusion associated with lower rates of help-seeking)."

In other words, this is pretty shitty. But hardly surprising.

Also not surprising: "Substantial economic costs are the consequence both for people with MDD and for society, because of low rates of treatment and recovery."

So, this is mostly more confirmation that a huge treatment gap exists with respect to depression. Part of the treatment gap revealed here is especially vexing. While over half of those with depression believe they need treatment and have access to it, only a small number of them actually seek it out, while an even smaller number adhere to it to the degree that it reaches the threshold of minimally adequate. The study notes that this can trace back to a few things, including the perceived quality of available services and high rates of treatment drop-out.

In all, we're talking about hundreds of millions of people with depression (there are 350 million in total worldwide) who are suffering when they may be helped by treatment. That shouldn't really be acceptable.

Finally, in a King's College statement, Thornicroft offers this: "Providing treatment at the scale required to treat all people with depression is crucial, not only for decreasing disability and death by suicide, but also from a moral and human rights perspective, and to help people to be fully productive members of society."

Get six of our favorite Motherboard stories every day by signing up for our newsletter.

from Most People With Depression Go Without Treatment

Someone Built a Working Four-Bit Computer Out of Cardboard and Marbles

Computers aren't magic. Underneath your fingertips right now are wires and logic gates. Switches. A 1 becomes a 0; two 1s become a 1; a 1 and a 0 become a 0; a 1 and a 0 become a 1; two 0s become a 1. Simply, different sorts of gates take on/off signals or pairs of on/off signals and emit more on/off signals in response. The most basic idea of logical computation and binary arithmetic goes back to Leibniz in 1705.

You don't need electricity to implement binary operations, though it certainly helps. All you really need is a way to ensure that somehow the binary inputs to the gates below (the As and Bs) result in the correct binary output (the X). When you have that, you have the materials for a computer and can theoretically do anything.

For those that remain doubtful, Github user lapinozz is here to the rescue with a 4-bit computer constructed out of cardboard and marbles. It's pretty amazing:

Here's an AND gate below. As you can see in the chart above, an AND gate is only supposed to output a 1 (a marble) if there are two 1 inputs. Otherwise, it outputs a 0 (no marble).

In a computer, logic gates are combined into constructs known as half-adders and full-adders. This is how logic becomes arithmetic. A half-adder looks like this:

So, the cardboard computer is actually a combination of adders, as below.

"I built it with my little sisters for a science activity, it can add numbers from 0 to 15 for a maximum computable number of 30," lapinozz writes. "We made it from scratch and at the time I didn’t see any of the various kind of calculator that have been made using Lego, wood and other, so it’s a completely new model!"

from Someone Built a Working Four-Bit Computer Out of Cardboard and Marbles

Neuroscientists Develop AI-Based Method for Hacking Fear

Imagine a library of fear. Lining its aisles aren't photo albums of snakes and plummeting aircraft, but are instead volumes containing cross-sectional scans of human brains. They are familiar fMRI images of folded grey matter and feature red-yellow flames of neural activity. In the fear library, these flames vary volume to volume and fear to fear. Different fears feature unique, consistent signatures in the brain, which is the fact that allows them to be so neatly cataloged.

This is a real thing: Distinct fears have distinct signatures. They can be identified by recurring patterns of activity. This is what allows for our library, but it also implies something else: that we might be able to use these patterns to manipulate fears.

This appears to indeed be the case, according to research published Monday in the journal Nature Human Behaviour courtesy of UCLA cognitive neuroscientist Hakwan Lau and colleagues at Columbia University and the Nara Institute of Science and Technology in Japan. Fears can be erased, much as they are in aversion therapy—where phobias are conditioned away through exposure to the subject of the phobia, an oft unpleasant process—but without every having to conjure the fear itself.

Image: Lau et al

The process behind the technique is called "decoded neurofeedback." In the group's setup, fears are first created by giving electric shocks to study participants while at the same time presenting them an image of a colored vertical line. Over time, the specific color of line corresponding to the shocks becomes "scary," and the associated brain activity is recorded. This activity is then fed into an artificial intelligence visual recognition algorithm, which abstracts a pattern.

The AI algorithm is then used to observe further brain activity for artifacts of that same pattern, which will appear in fragmented form even though the participant is resting and is not currently being subjected to the fear stimulus (the colored line).

Once the algorithm has its fear patterns acquired, they can be used to deprogram the specific fears by associating positive rewards with the fear-associated colored lines. "Whenever your brain is representing or 'thinking about' the red line, one of the scary things, we [tell the subject], 'Congratulations you won 10 cents,'" Lau explained in an interview. "So, now whenever the red thing happens, instead of being paired with the electric shock, now it's associated with a positive monetary reward."

"With this procedure we can condition out the fear," he said.

This sounds like pretty standard psychology, but the key thing is that the subject never has to consciously think about the scary thing for it work. The recognition algorithm just has to see a fleeting fragment of the fear memory to trigger a reconditioning stimulus (passing out money). After the subjects were "reprogrammed," they were again shown images of the once-scary lines. Their fearfulness, as represented by standard skin sweat fear responses, had diminished.

The obvious catch is that somehow the original fear neural signature has to be acquired at some point for the AI to recognize it and trigger reprogramming reward events. This brings us back to our fear library. What Lau and colleagues have found is that these fear patterns are actually shared across many different people. There appear to be general fear fingerprints corresponding to different phobias that occur across populations. Opening the volume for arachnophobia in our library would then reveal just a single neural pattern. As it turns out, fears can be inferred from fMRI scans with up to 80 percent accuracy. This part of the group's research is so-far unpublished, but Lau expects that a new paper will be out within a few months.

"Using other people, we can infer what your brain's spider pattern will be with up to 80 percent [accuracy]."

"The idea is that you see dogs, cats, oranges, and butterflies, and then I see the same thing, " Lau explains. "Oranges, dogs, cats, and butterflies. There's a way to calibrate our brains' patterns in the same space. Once our brains are calibrated, you don't have to see spiders anymore [to capture the associated brain activity]. I can go see spiders. I can watch spiders for hours. Then, I know my pattern for spiders, and I can infer your spider pattern. It sounds sci-fi but it can be done, and it can be done for more than one brain. Using other people, we can infer what your brain's spider pattern will be with up to 80 percent [accuracy]."

Lau assures that this doesn't work the other way. We can't insert a fear or negative association into the brain through this method—only positive associations. Fears can be subtracted and not added. So, the consequences of getting it wrong are only that the patient might wind up with an artificial positive association.

There are some remaining barriers before we might see this in a clinical setting. For one thing, it's uncertain how long the reprogramming lasts. There's also a well-known phenomenon in PTSD in which fears can re-emerge as individuals encounter new contexts. If, say, someone gets rid of their fear of cars following a car accident, they may re-experience PTSD fear if they return to the location of the accident. This may happen here as well, which will only be revealed when similar experiments are run that swap simple colored lines for richer real-world stimuli.

from Neuroscientists Develop AI-Based Method for Hacking Fear

New Antibody Neutralizes Nearly Every HIV Strain

Researchers at the National Institutes of Health have isolated an antibody from an HIV-positive patient that is capable of neutralizing 98 percent of HIV strains. It's a success rate that makes it easily the most potent, wide-reaching HIV antibody and one that may have profound implications for the treatment and prevention of the disease. The group's work is published in the current issue of the journal Immunity.

Antibodies are great would-be weapons against HIV. An antibody is, generally, a protein produced by the immune system that tries to neutralize pathogens like viruses and bacteria. It works by binding chemically to the invader, which it may "tag" for further attack by the immune system or it may block off one of the pathogen's mechanisms—picture quasi-biological molecular "spikes"—for invading a friendly cell.

The latter trick is crucial for fighting viruses. That's the whole existence of a virus—binding to a healthy cell and then taking it over by injecting in some new genetic material. If an antibody is attached to a binding site used by a virus to hijack a human cell, then the virus is rendered impotent. It's a great defense, in theory.

In reality, there are a lot of different HIV strains and a lot of different sites to bind to. The virus is always changing and evolving. Some patients, however, wind up producing broadly-neutralizing antibodies, which are able to hit a large number of strains. As such, they can potentially do real damage against a wholesale HIV infection made up of diverse and mutating viral strains.

Image: interactive-biology

The prior "record" for HIV neutralization was 90 percent. That came courtesy of an antibody known as VRC01, which is currently be assessed for clinical applications. The new antibody, known as N6, is so effective because of its targeting of a region (known as V5) found on the HIV viruses exterior envelope that is very often conserved across the pathogen's evolution—it remains constant, relatively.

N6 has a couple of other advantages. One, it's particularly effective against HIV strains that have become resistant to antibodies in the same antibody class. Two (but very related to the above), it's structurally built to evade a common defensive mechanism wielded by the HIV virus called a steric clash.

To be clear, we're not really talking cures here. The most immediate application would be in vaccine development, but it will likely have implications for treatment as well. In any case, N6 faces years of further testing. For perspective, the aforementioned VRC01 was discovered all the way back in 2010 and it's still in clinical trials.

from New Antibody Neutralizes Nearly Every HIV Strain