ArchieBunker writes: In the streaming era, music is accessed from a variety of online services, ephemeral in nature and never living on board the device. However, the online audio revolution really kicked off with the development of one very special format. The subject of bitter raps and groundbreaking lawsuits, this development from Germany transformed the music industry as we know it. Twenty-five years on from the date the famous '.mp3' filename was chosen, HackADay takes a look back at how it came to be, and why it took over the world.

Elon Musk says Neuralink recipients will be able to stream music directly into their brains, with no headphones needed. Futurism reports: It's still entirely unclear if Neuralink's brain-computer interfaces will ever be able to bypass the ear, the cochlear nerves, and beam music magically to the auditory brain. But Musk has already set his sights on bigger things. When asked by another Twitter user if Neuralink's devices could also help "estimulate the release of oxytocin, serotonin, and other chemicals when needed," Musk replied with another "yes." To make that feature and more a reality, Neuralink is looking to add more, well, brains to its operations. "If you've solved hard problems with phones/ wearables (sealing, signal processing, inductive charging, power management, etc.), please consider working at [Neuralink]," Musk tweeted over the weekend. On August 28th, the mysterious computer-brain interface startup will release an update on its progress.

An anonymous reader quotes a report from The New York Times: Researchers in Singapore have developed an apparatus that can be placed in a window to reduce incoming sound by 10 decibels. The system was created by a team of scientists, including Masaharu Nishimura, who came up with the basic concept, and Bhan Lam, a researcher at Nanyang Technological University in Singapore. Their results were published on Thursday in Scientific Reports. The prototype is not yet the most practical device in real world conditions, but it points the way toward the development of technologies that may help ease the strain of noisy city living.

Borrowing from the same technological principles used in noise-canceling headphones, the team expanded the concept to fit an entire room by placing 24 small speakers in a window. The speakers emit sound waves that correspond to the incoming racket and neutralize it -- or, at least some of it. The system is based on the frequency of the sound waves and, for now, the optimal range is between 300 and 1,000 hertz. [...] The system uses a microphone outside the window to detect the repeating sound waves of the offending noise source, which is registered by a computer controller. That in turn deciphers the proper wave frequency needed to neutralize the sound, which is transmitted to the array of speakers on the inside of the window frame. The speakers then emit the proper "anti" waves, which cancel out the incoming waves, and there you have it: near blissful silence. Unfortunately, there are some limitations. The system works best from the types of steady noise sources found within the optimal frequency range and isn't great at neutralizing sporadic noises. Also, since human voices don't fit within most of that range, they won't be canceled out.

Researchers at Zhejiang University and Microsoft claim they've developed an AI system -- DeepSinger -- that can generate singing voices in multiple languages by training on data from music websites. From a report: In a paper published on the preprint Arxiv.org, they describe the novel approach, which leverages a specially-designed component to capture the timbre of singers from noisy singing data. The work -- like OpenAI's music-generating Jukebox AI -- has obvious commercial implications. Music artists are often pulled in for pick-up sessions to address mistakes, changes, or additions after a recording finishes. AI-assisted voice synthesis could eliminate the need for these, saving time and money on the part of the singers' employers.

But there's a darker side: It could also be used to create deepfakes that stand in for musicians, making it seem as though they sang lyrics they never did (or put them out of work). In what could be a sign of legal battles to come, Jay-Z's Roc Nation label recently filed copyright notices against videos that used AI to make him rap Billy Joel's "We Didn't Start the Fire." As the researchers explain, singing voices have more complicated patterns and rhythms than normal speaking voices. Synthesizing them requires information to control the duration and the pitch, which makes the task challenging. Plus, there aren't many publicly available singing training data sets, and songs used in training must be manually analyzed at the lyrics and audio level.

America's studios, creators and marketers are relying, more than ever, on digital platforms that allow them to gauge what audiences like-- and would like to see more of. From a report: They're not just looking for test screenings, either. They're looking to check in with potential audiences at every stage of production, from before a script is written until the moment their new TV show, film or music video debuts. Ever since George Lucas ushered in the era of endless sequels (and prequels), Hollywood executives have tried to capitalize on the success of the Last Big Thing by churning out more of it. But content budgets are increasing far faster than established franchises can keep up. Netflix is projected to spend more on new and acquired content in 2020 -- $17 billion -- than Apple spent on research and development in 2019. With stakes that high, minimizing risk when creating new content "at scale" means treating it like any other mass-market product. Executives, producers, writers, directors and marketers need to be able to consistently craft programs that are more likely than not to find their target audiences. Critical approval and industry awards -- even box-office blowouts -- while nice, aren't the endgame for most.

navindra (Slashdot reader #7,571) is a professional coder and an amateur musician. He writes: Music and technology have always been intertwined and this has never been truer with computer technology. I set out on a simple quest to find a better metronome, settled for a digital sample sequencer, but ended up fascinated by the Organelle music computer running Pure Data on a Raspberry Pi and Linux.
"The entire system runs open source software and may be customized at every level," explains its web page. From navindra's article: [T]he Organelle is hard to define, but it is nothing less than a music computer, almost demanding that one ascend from mere musician to virtual instrument designer.

Organelle patches are readily available for beat-making purposes, including sample sequencer or drum machine functions, multi-dimensional sequencers, and, yes, even a straight up metronome. Beyond the factory patches, the user community contribute their own creations... Pure Data code initially looks like low-level symbolic code — fortunately, it is easy to make something musical in short order, if one has a handle on sound synthesis and MIDI concepts.

The Organelle runs Linux on a Raspberry Pi with everything that entails — it has a boot up time of 12 seconds, yields a 6 hour battery life on 4 AA batteries (roughly proportional to the volca power draw), and things like swapping samples is trivially done with a file manager, eliminating the need for special software. Similarly, adding or modifying patches is just as easy — modifying the existing Metronome patch with a new sound could be as easy as dropping in a new .wav file.

Needless to say, the Organelle is constantly evolving and improving...

"Swirls of color and images blend together as faces, scenery, objects, and architecture transform to music." That's how AI training company Lionbridge is describing Neural Synesthesia.

Slashdot reader shirappu explains: Neural Synesthesia is an AI art project that creator Xander Steenbrugge calls a collaboration between man and machine. To create each piece, he feeds a generative network with curated image datasets and combines the ever-transforming results with music that is programmed to control the shifting visuals.
Steenbrugge describes how the music controls the visuals in an interview with Lionbridge: I think coding for the first rendered video took over six months because I was doing it in my spare time. The biggest challenge was how to manipulate the generative adversarial network (GAN)'s latent input space using features extracted from the audio track. I wanted to create a satisfying match between visual and auditory perception for viewers.

I apply a Fourier Transform to extract time varying frequency components from the audio. I also perform harmonic/percussive decomposition, which basically separates the melody from the rhythmic components of the track. These three signals (instantaneous frequency content, melodic energy, and beats) are then combined to manipulate the GANs latent space, resulting in visuals that are directly controlled by the audio...

[Y]ou are not limited by your own imagination. There's an entirely alien system that is also influencing the same space of ideas, often in unexpected and interesting ways. This leads you as a creator into areas you never would have wandered by yourself.

Bytedance said it's evaluating changes to the corporate structure of its TikTok business as U.S. concerns grow over the parent company's Chinese origins. Bloomberg reports: Executives are discussing options such as creating a new management board for TikTok and establishing a separate headquarters for the app outside of China to distance its operations from Beijing, according to a person familiar with the discussions. Short-video and music app TikTok currently doesn't have its own headquarters separate from Bytedance, which was founded in China and is incorporated in the Cayman Islands. TikTok is considering a number of locations for a global base, the person familiar with the plans said, asking not to be named discussing information that's not public. Its five largest offices are in Los Angeles, New York, London, Dublin and Singapore. "We will move forward in the best interest of our users, employees, artists, creators, partners, and policy makers," TikTok said in a statement.

An anonymous reader quotes a report from Phys.Org: A ringing bell vibrates simultaneously at a low-pitched fundamental tone and at many higher-pitched overtones, producing a pleasant musical sound. A recent study, just published in the Journal of the Atmospheric Sciences by scientists at Kyoto University and the University of Hawai'i at Mnoa, shows that the Earth's entire atmosphere vibrates in an analogous manner, in a striking confirmation of theories developed by physicists over the last two centuries. In the case of the atmosphere, the "music" comes not as a sound we could hear, but in the form of large-scale waves of atmospheric pressure spanning the globe and traveling around the equator, some moving east-to-west and others west-to-east. Each of these waves is a resonant vibration of the global atmosphere, analogous to one of the resonant pitches of a bell.

Now in a new study by Takatoshi Sakazaki, an assistant professor at the Kyoto University Graduate School of Science, and Kevin Hamilton, an Emeritus Professor in the Department of Atmospheric Sciences and the International Pacific Research Center at the University of Hawaii at Mnoa, the authors present a detailed analysis of observed atmospheric pressure over the globe every hour for 38 years. The results clearly revealed the presence of dozens of the predicted wave modes. The study focused particularly on waves with periods between 2 hours and 33 hours which travel horizontally through the atmosphere, moving around the globe at great speeds (exceeding 700 miles per hour). This sets up a characteristic "chequerboard" pattern of high and low pressure associated with these waves as they propagate. "For these rapidly moving wave modes, our observed frequencies and global patterns match those theoretically predicted very well," stated lead author Sakazaki. "It is exciting to see the vision of Laplace and other pioneering physicists so completely validated after two centuries."

Walmart said this week that it was converting some of its parking lots into drive-in theaters for the summer as the movie industry struggles amid the coronavirus pandemic. From a report: The retail behemoth is converting 160 of its parking lots across the US into drive-ins. These theaters will open in early August and remain open through October. The Walmart Drive-In will feature movies programmed by Tribeca Enterprises, the company behind the Tribeca Film Festival, which recently launched a summer movie drive-in series bringing films, music, and sporting events to as many US drive-ins as possible.

Facebook confirmed that it's shutting down two of its little-known social media apps shortly after their launch. TikTok rival Lasso and Pinterest rival Hobbi will both be terminated on July 10. CNBC reports: Lasso, the more popular of the two, allowed people to record videos up to 15 seconds long and overlay music on top. It was launched a year-and-a-half ago. Lasso issued a push alert to users on Wednesday telling them that it will be shutting down. In order to try to compete with ByteDance's TikTok, which has been downloaded over 2 billion times, Facebook-owned Instagram has developed its own video-music mix feature called "Reels."

Pinterest-rival Hobbi, which only went live on the Apple App Store in the U.S. in February, also issued a push alert to users on Wednesday letting them know that it was closing down. The app, designed by Facebook's New Product Experimentation team, allows people to document, share and organize their hobbies. It has received just 7,000 downloads, according to Sensor Tower.

An anonymous reader quotes a report from Wired: The threat of ransomware may seem ubiquitous, but there haven't been too many strains tailored specifically to infect Apple's Mac computers since the first full-fledged Mac ransomware surfaced only four years ago. So when Dinesh Devadoss, a malware researcher at the firm K7 Lab, published findings on Tuesday about a new example of Mac ransomware, that fact alone was significant. It turns out, though, that the malware, which researchers are now calling ThiefQuest, gets more interesting from there. In addition to ransomware, ThiefQuest has a whole other set of spyware capabilities that allow it to exfiltrate files from an infected computer, search the system for passwords and cryptocurrency wallet data, and run a robust keylogger to grab passwords, credit card numbers, or other financial information as a user types it in. The spyware component also lurks persistently as a backdoor on infected devices, meaning it sticks around even after a computer reboots, and could be used as a launchpad for additional, or "second stage," attacks. Given that ransomware is so rare on Macs to begin with, this one-two punch is especially noteworthy.

Though ThiefQuest is packed with menacing features, it's unlikely to infect your Mac anytime soon unless you download pirated, unvetted software. Thomas Reed, director of Mac and mobile platforms at the security firm Malwarebytes, found that ThiefQuest is being distributed on torrent sites bundled with name-brand software, like the security application Little Snitch, DJ software Mixed In Key, and music production platform Ableton. K7's Devadoss notes that the malware itself is designed to look like a "Google Software Update program." So far, though, the researchers say that it doesn't seem to have a significant number of downloads, and no one has paid a ransom to the Bitcoin address the attackers provide. [...] Given that the malware is being distributed through torrents, seems to focus on stealing money, and still has some kinks, the researchers say it was likely created by criminal hackers rather than nation state spies looking to conduct espionage.

Max Weinbach, writing for Android Police: I have been using Samsung phones every day for almost 4 years. It was because Samsung had fantastic hardware paired with --depending on the year -- good software. 2020 is the first year in a while I'm not using a Samsung phone as my daily driver. The reason? Ads. Ads in Samsung phones never really bothered me, at least not until the past few months. It started with the Galaxy Z Flip. A tweet from Todd Haselton of CNBC is what really caught my eye. Samsung had put an ad from DirectTV in the stock dialer app. This is really something I never would have expected from any smartphone company, let alone Samsung. It showed up in the "Places" tab in the dialer app, which is in partnership with Yelp and lets you search for different businesses directly from the dialer app so you don't need to Google somewhere to find the address or phone number. I looked into it, to see if this was maybe a mistake on Yelp's part, accidentally displaying an ad where it shouldn't have, but nope. The ad was placed by Samsung, in an area where it could blend in so they could make money.

Similar ads exist throughout a bunch of Samsung apps. Samsung Music has ads that look like another track in your library. Samsung Health and Samsung Pay have banners for promotional ads. The stock weather app has ads that look like they could be news. There is also more often very blatant advertising in most of these apps as well. Samsung Music will give you a popup ad for Sirius XM, even though Spotify is built into the Samsung Music app. You can hide the SiriusXM popup, but only for 7 days at a time. A week later, it will be right back there waiting for you. Samsung will also give you push notification ads for new products from Bixby, Samsung Pay, and Samsung Push Service.

The Apple executive in charge of the App Store in Europe said that the company's policies ensure a level playing field for developers and ease-of-use for customers as regulatory scrutiny over the platform mounts. From a report: "Our efforts to help developers succeed are broad, deep and ongoing, and they extend to apps -- in music, email, or a variety of other categories -- that compete with some aspect of our business," Daniel Matray, the iPhone maker's head of App Store and media services in Europe, said in a speech Tuesday at a four-day virtual conference hosted by Forum Europe. The speech comes as Apple faces antitrust probes in the European Union and U.S. over rules it imposes on developers. In particular, regulators are taking aim at the requirement that apps use the company's in-house payment service, which takes a cut of 15% to 30% of most subscriptions and in-app purchases. Matray said that about 85% of apps it hosts don't pay Apple a commission because they're free or earn revenue through other means. Further reading: How Apple Stacked the App Store With Its Own Products.

CNET is celebrating its 25th anniversary with articles remembering the 1990s — including that moment "when Hollywood finally noticed the web," calling it "a flawed but fun snapshot of the moment the internet took over the world..."

"Twenty-five years ago, cinema met cyberspace in a riot of funky fashion, cool music and surveillance paranoia. It began in May 1995 with the release of Johnny Mnemonic, a delirious sci-fi action dystopia matching Keanu Reeves with seminal cyberpunk author William Gibson. In July, Sandra Bullock had her identity erased in conspiracy thriller The Net. In August, Denzel Washington pursued Russell Crowe's computer-generated serial killer in Virtuosity, and in September Angelina Jolie found her breakthrough role in anarchic adventure Hackers. In October, Kathryn Bigelow served up dystopian thriller Strange Days. It's hard to know what's most dated about these mid-'90s curios: the primitive-looking effects, the funky fashions or the clunky technology depicted on screen. But now, 25 years later, they've proved prescient in their concerns about surveillance, corporate power and the corruption of what seemed to be an excitingly democratic new age...

Most tellingly, Johnny Mnemonic and the other tech-focused films of 1995 all express fears around the misuse of surveillance in a connected world. The Net updates the paranoia of '70s thrillers The Conversation and The Anderson Tapes, and each movie features an unholy alliance of avaricious corporate bad guys and authoritarian law enforcement. Or as Matthew Lillard's character puts it in Hackers, "Orwell is here and livin' large!"

But the whistleblowing heroes of Hackers, The Net and Johnny Mnemonic use their skills to subvert and unpick the establishment's grip on technology. Hackers in particular radiates an infectious idealism as the diverse crew of anarchic youngsters rollerblade rings around the greedy suits and clueless cops, "snooping onto them as they snoop onto us". The movie highlights technology's potential to be a tool for wrongdoing and a democratic, open medium where you can be who you want to be... Sadly, 1995's wave of technology-themed movies have one other thing in common. They all bombed.
CNET's reporter gets new quotes from the director of Hackers — as well as one of that film's then-15-year-old technical advisors, Nicholas Jareck. "For all its exaggerations," he says, "it does a decent job of showing the hacker spirit — those kids were tinkerers, experimenting, reveling in their ability to figure something out. It's a celebration of human ingenuity."

Johnny Mnemonic. "Speaking on the phone from New York, Longo's memories are peppered with entertaining asides about who was 'evil,' 'a dick,' 'an idiot' or 'a fucking idiot.'"

A high school computer science teacher claims there's an "unacknowledged failure" of America's computer science (CS) classes at the high school and junior high school level. "Visit classrooms and you'll find students working with robotic sensors, writing games and animations in Scratch, interfacing with Arduino microcontrollers, constructing websites, and building apps with MIT App Inventor...

"Look underneath the celebratory and self-congratulatory remarks, however, and you'll find that, although contemporary secondary education is quite good at generating initial student interest, it has had much less success at sustaining that engagement beyond a few weeks or months, and has frankly been ineffectual in terms of (a) measurable learning for the majority of students; (b) boosting the number of students who take a second CS course, either in high school or college; and (c) adequately preparing students for CS college study."

Long-time Slashdot reader theodp writes: In " A New Pedagogy to Address the Unacknowledged Failure of American Secondary CS Education ," high school computer science teacher Scott Portnoff argues that a big part of the problem is the survey nature of today's most popular high school CS course offerings — Exploring Computer Science (ECS) and AP Computer Science Principles (AP CSP) — both of whose foundational premise is that programming is just one of many CS topics. "Up until a decade ago," Portnoff explains, "introductory high school computer science classes were synonymous with programming instruction, period. No longer."

This new status quo in secondary CS education, Portnoff argues, resulted from baseless speculation that programming was what made Java-based AP CS A inaccessible, opposed to, say, an uninspiring or pedagogically ineffective version of that particular curriculum, or a poorly prepared instructor. It's quite a departure from the 2011 CSTA K-12 Computer Science Standards, which made the case for the centrality of programming in CS education ("Pedagogically, computer programming has the same relation to studying computer science as playing an instrument does to studying music or painting does to studying art. In each case, even a small amount of hands-on experience adds immensely to life-long appreciation and understanding").
This teacher believes that programming languages are acquired rather than learned, just like any other human language — and concludes the solution is multi-year courses focused on one programming language until proficiency is fully acquired.

For this reason, for the last seven years he's also been making his students memorize small programs, and then type them out perfectly, arguing that "the brain subconsciously constructs an internal mental representation of the syntax rules implicitly by induction from the patterns in the data."

Michael Hawley, a computer programmer, professor, musician, speechwriter and impresario who helped lay the intellectual groundwork for what is now called the Internet of Things, died on Wednesday at his home in Cambridge, Mass. He was 58. From a report: The cause was colon cancer, said his father, George Hawley. Mr. Hawley began his career as a video game programmer at Lucasfilm, the company created by the "Star Wars" director George Lucas. He spent his last 15 years curating the Entertainment Gathering, or EG, a conference dedicated to new ideas. In between, he worked at NeXT, the influential computer company founded by Steve Jobs after he left Apple in the mid-1980s, and spent nine years as a professor at the Massachusetts Institute of Technology Media Lab, a seminal effort to push science and technology into art and other disciplines. He was known as a scholar whose ideas, skills and friendships spanned an unusually wide range of fields, from mountain climbing to watchmaking. Mr. Hawley lived with both Mr. Jobs and the artificial intelligence pioneer Marvin Minsky, published the world's largest book, won first prize in an international competition of amateur pianists, played alongside the cellist Yo-Yo Ma at the wedding of the celebrity scientist Bill Nye, joined one of the first scientific expeditions to Mount Everest, and wrote commencement speeches for both Mr. Jobs and the Google co-founder Larry Page.

Two of Mr. Hawley's Media Lab projects -- Things That Think and Toys of Tomorrow -- anticipated the Internet of Things movement, which aims to weave digital technology into everything from cars to televisions to home lighting systems. Led by companies like Amazon, Google, Intel and Microsoft, the movement is now a $248 billion market, according to the market research firm Statista. Mr. Hawley developed "a pattern of ideas that emerged long before the Internet of Things," Nicholas Negroponte, founder of the Media Lab, said in an email. "I would call that pattern not artificial intelligence, but intelligence in the artificial," he wrote. Mark Seiden, an independent computer security consultant who met Mr. Hawley in the early 1980s when they were both working at IRCAM, a music lab in Paris, and eventually hired him at Lucasfilm, compared Mr. Hawley's exploits to those of George Plimpton, the writer whose participatory kind of journalism had him masquerading as a boxer, a professional football player, a circus performer and a stand-up comedian.

Amazon is looking to add 24/7 live programming to its Prime Video service, according to Protocol. The new channels could include live news, music and sports as well as scheduled movies and TV show showings. From the report: Speaking under the condition of anonymity, an industry insider told Protocol that Amazon has been "actively pursuing" deals to license live and linear programming. "You should assume they're talking to everybody," he said. By adding live programming to Prime Video, Amazon could differentiate itself from services like Netflix and Disney+ that are focused exclusively on on-demand video. The move is also a response to the growing popularity of linear streaming services like Pluto and Xumo, and ultimately could be part of a different take on live TV: Instead of licensing the same costly programming bundles as traditional cable services, Amazon may be looking to combine its existing on-demand content and a much more narrow take on must-see live TV.

Amazon has been experimenting with live programming for Prime Video over the past several years, which included licensing NFL Thursday Night Football as well as the English Premier League. In the future, it may also stream live concerts, political debates and news programming, according to a job listing for Amazon's Prime Video live events team. "This is a transformative opportunity, the chance to be at the vanguard of a program that will revolutionize Prime Video," that job listing reads. Beyond individual live events, Amazon is also looking to license complete 24/7 feeds. "Linear TV enables customers to watch 24/7 streams of their favorite TV stations airing programs including sports, news, movies, award shows, special events and TV shows," one job listing details. Another specifically singles out live broadcasters and cable networks as potential partners.

Optimism about the internet's role in politics peaked around the time of the Arab Spring, then steadily collapsed into alarm and despair until this weekend, when it ticked up again after President Donald Trump held a disappointing campaign rally. From a report: There are various ways to interpret the lower-than-expected turnout at the Tulsa, Oklahoma event, but among the most intriguing was the claim from a group of Korean pop fans that they'd undercut the campaign by coordinating to reserve thousands of tickets, then not showing up. They are likely giving themselves too much credit. Still, the narrative took hold for online observers as an example of a rare bright spot in the social media hellscape. The surge in activism from young Korean pop music enthusiasts has been one of the stranger plot lines of a uniquely unsettled time in American politics. Working together, they've rendered Twitter hashtags like #WhiteLivesMatter useless by filling them with music video clips, and they crashed a mobile app established by the Dallas Police Department to collect evidence of illegal activity at protests by overwhelming it with data.

This has gripped the imagination of some internet commentators, who noted how young people have reconstituted their "lightning-fast coordination and prodigious spamming abilities" for what the fans believe are righteous political causes. But spamming has historically been seen as a bad thing. When right-wing trolls coordinate to do things like pollute hashtags, pile onto people they dislike or disrupt the process of government it's regularly described as a serious threat to democracy. The tactics are remarkably similar, though the end goals are different.

IBM's research blog shares an article about "polyphonic music prediction using the Johann Sebastian Bach chorales dataset" achieved by using "biologically plausible neurons," a new approach to deep learning "that incorporates biologically-inspired neural dynamics and enables in-memory acceleration, bringing it closer to the way in which the human brain works." At IBM Research Europe we have been investigating both Spiking Neural Networks (SNNs) and Artificial Neural Networks (ANNs) for more than a decade, and one day we were struck with the thought: "Could we combine the characteristics of the neural dynamics of a spiking neuron and an ANN?" The answer is yes, we could. More specifically, we have modelled a spiking neuron using a construct comprising two recurrently-connected artificial neurons — we call it a spiking neural unit (SNU)... It enables a reuse of architectures, frameworks, training algorithms and infrastructure. From a theoretical perspective, the unique biologically-realistic dynamics of SNNs become available for the deep learning community...

Furthermore, a spiking neural unit lends itself to efficient implementation in artificial neural network accelerators and is particularly well-suited for applications using in-memory computing. In-memory computing is a promising new approach for AI hardware that takes inspiration from the architecture of the brain, in which memory and computations are combined in the neurons. In-memory computing avoids the energy cost of shuffling data back and forth between separate memory and processors by performing computations in memory — phase change memory technology is a promising candidate for such implementation, which is well understood and is on its way to commercialization in the coming years. Our work involves experimental demonstration of in-memory spiking neural unit implementation that exhibits a robustness to hardware imperfections that is superior to that of other state-of-the-art artificial neural network units...

The task of polyphonic music prediction on the Johann Sebastian Bach dataset was to predict at each time step the set of notes, i.e. a chord, to be played in the consecutive time step. We used an SNU-based architecture with an output layer of sigmoidal neurons that allows a direct comparison of the obtained loss values to these from ANNs. The SNU-based network achieved an average loss of 8.72 and set the SNN state-of-the-art performance for the Bach chorales dataset. An sSNU-based network further reduced the average loss to 8.39 and surpassed corresponding architectures using state-of-the-art ANN units.
Slashdot reader IBMResearch notes that besides being energy-efficient, the results "point towards the broad adoption of more biologically-realistic deep learning for applications in artificial intelligence."