Tech Talk iconTech Talk

Use a GPU to Turn a PC Into a Supercomputer

As Moore’s Law slows for CPUs, dedicated graphics co-processors are picking up some of the slack. Just as GPUs are changing the game in deep learning and autonomous cars, the GPU-powered desktop PC might even begin to keep pace with the conventional supercomputer for a portion of supercomputer applications. 

For instance, a group of Russian scientists are reporting this month that they’ve been able to solve computational problems in nuclear physics using an off-the-shelf, high-end PC containing a GPU. And, they say, after fine-tuning their algorithm for GPUs, they were able to run their calculations faster than the traditional, CPU-powered supercomputer their colleagues use. Bonus: they ran those calculations for free as opposed to their colleagues, who must pay for access to the supercomputer to run their computations.

Read More

Your Smart Watch Can Steal Your ATM PIN

Mobile systems and cyber security expert Yan Wang doesn’t wear a smart watch.

“It knows too much,” says Wang, an assistant professor of computer science at Binghamton University in Upstate New York. “If you are using a smart watch, you need to be cautious.”

Wearable devices can give away your PIN number, according to research by professor Yingying Chen at Stevens Institute of Technology and three of her current and former graduate students including Wang. By combining smart watch sensor data with an algorithm to infer key entry sequences from even the smallest of hand movements, the team was able to crack private ATM PINs with 80 percent accuracy on the first try and more than 90 percent accuracy after three tries.

“This was surprising even to those of us already working in this area,” Chen, who led the research, said in a press release. “It may be easier than we think for criminals to obtain secret and private information from our wearables by using the right techniques.” 

Read More
Photos: University of Washington

One Million Faces Challenge Even the Best Facial Recognition Algorithms

Helen of Troy may have had the face that launched a thousand ships, but even the best facial recognition algorithms may have had trouble finding her face in a crowd of one million strangers. The first benchmark test based on one million faces has shown how facial recognition algorithms from Google and other research groups around the world can still fall short in accurately identifying and verifying faces.

Facial recognition algorithms that had previously performed with more than 95 percent accuracy on a popular benchmark test involving 13,000 faces saw significant drops in accuracy when faced with the new MegaFace Challenge  involving one million faces. The best performer on one test, Google’s FaceNet algorithm, dropped from near-perfect accuracy on five-figure datasets to 75 percent on the million-face test. Other top algorithms dropped from above 90-percent accuracy on the small datasets to below 60 percent on the MegaFace Challenge. Some algorithms made the proper identification as seldom as 35 percent of the time.

“Megaface's key idea is that algorithms should be evaluated at large scale,” says Ira Kemelmacher-Shlizerman, an assistant professor of computer science at the University of Washington in Seattle and the project’s principal investigator. “And we make a number of discoveries that are only possible when evaluating at scale.”

Read More
Thomas 'Tom' Wheeler, chairman of the U.S. Federal Communications Commission (FCC).

More 5G Spectrum Coming Soon, FCC Chair Says

The United States will rush to make high-frequency spectrum bands available for emerging 5G technologies, promised Tom Wheeler, Chairman of the Federal Communications Commission, in an address on Monday. Wheeler said he will share his plan for opening these bands on Thursday, with an official FCC vote scheduled for 14 July.

Read More
A woman attaches GPS buoys to wooden sea turtle replicas.

Dead Turtles Equipped with GPS Trackers to Solve the Mystery of Their Own Murders

Hundreds of dead sea turtles wash up on the shores of Chesapeake Bay each year. Strangely, these events known as strandings seem to occur more often along certain stretches of shoreline. But even while so many turtles wind up in these spots, no one knows exactly where they’re coming from or how most of them died.  

To try to figure it out, a master’s student is setting a handful of GPS-equipped turtle carcasses afloat in Chesapeake Bay this summer. Bianca Santos, a graduate student at William and Mary’s School of Marine Science at the Virginia Institute of Marine Science, released two so-called “drifters” from a boat last week and hopes to launch up to 10 this season.

She hopes that by tracking how the carcasses are pushed along by wind and ocean currents, she can figure out where those hundreds of beached turtles floated in from, and whether some spots in the ocean might prove more deadly to turtles than others.

Read More

China's New Supercomputer Is World's Most Powerful

For the last three years, China has topped the Top500 list of the most powerful supercomputers with its massive Tianhe-2. But today, the Top500 group announced that Tianhe-2 has been ousted by another Chinese supercomputer, the Sunway TaihuLight. The new machine, which is based at the National Supercomputing Center in Wuxi, can perform a key benchmark test called Linpack at 93 petaflops (a thousand trillion floating point operations per second)—nearly three times the speed of the Tianhe-2. 

The new rankings further solidify China’s status as a supercomputing force to be reckoned with. In addition to this new machine, the United States has, for the first time, lost its status as the country with the most systems on the list; China now has 167 systems to the U.S.’s 165.

Unlike the Tianhe-2, which used Intel Xeon chips to take the top spot, the processors inside the Sunway TaihuLight are homegrown. At each of the machine’s 40,960 nodes, the supercomputer uses a new 260-core chip, designed by the Shanghai High Performance IC Design Center.

According to the Top500 site, Sunway TaihuLight will be used for research and engineering work, including weather modeling and advanced manufacturing.

Although supercomputing progress has slowed in recent years, there are still-more-powerful machines on the horizon. The United States, for one, has a batch of new machines in the works. According to a report on Sunway TaihuLight by Top500 team member Jack Dongarra, 2018 could see the arrival of three new U.S. Department of Energy machines, the speediest of which will be Summit, a 200-petaflop supercomputer to be installed at Oak Ridge National Laboratory in Tennessee.

The top 10 supercomputers from the June 2016 list.
Name Country Teraflops Power (kW)
Sunway TaihuLight China 93,015 15,371
Tianhe-2 China 33,863 17,808
Titan United States 17,590 8,209
Sequoia United States 17,173 7890
K Computer Japan 10,510 12,660
Mira United States 8,587 3,945
Trinity United States 8,101 N/A
Piz Daint Switzerland 6,271 2,325
Hazel Hen Germany 5,640 N/A
Shaheen II Saudi Arabia 5,537 2,834

Illustration by Getty Images

DAO May Be Dead After $60 Million Theft

In the world of cryptocurrency, seven-figure heists are a rite of passage. And today, Ethereum, a much-hyped blockchain currency and autonomous software platform, has come of age. This morning, participants in a lavishly-funded investment vehicle called The DAO woke up to an onslaught of alarm bells when it was discovered that a hacker had utilized a vulnerability in the code to drain the fund. 

At 4 a.m., Griff Green, a developer for, a company based in Germany that is building on the Ethereum blockchain and created The DAO, sent out this alert on the community’s Slack channel:

The DAO is being attacked. It has been going on for 3-4 hours, it is draining ETH at a rapid rate. This is not a drill.”

By midday, the attacker had stolen over 3 million Ether (Ethereum’s native currency), which at the time was worth more than 60 million U.S. dollars.

The DAO is a public investment fund that exists as a bundle of software on the Ethereum network. In a sale this spring, participants signed up for the fund by using their Ether to buy DAO tokens. This, in turn, gave them proportional ownership as well as the right to vote on investment proposals. The DAO software was intended to autonomously coordinate and enforce voting and fund allocation, thereby creating an investment vehicle that could operate without third-parties. The sale was much more successful than anyone expected, and by the end of May, Ether holders had dumped more than $150 million worth of their cryotocurrency into The DAO.

A day before the fund opened for business, a group of researchers identified critical flaws and biases in the process whereby participants vote on spending proposals. There was an appeal for a moratorium until fixes could be made, but today’s losses were caused by a bug that had gone unnoticed until now.

The DAO software gives token holders the ability to take their contributions and split from the larger group into their own identical DAOs. Once the split occurs, the person who created the new DAO has full control over what happens to the tokens, and after 27 days he or she is free to sell the tokens on an exchange to recoup the investment. People who choose this route should be able to leave with only as many tokens as they rightfully own. But late last night, someone found a way to split off with more than their fair share. According to sources, the hacker who did this now controls a satellite DAO that contains 100 times the amount of tokens that he initially invested.

Already, Ethereum developers are proposing ways to fix the situation. The easiest, and least controversial solution is for the network to adopt a new version of the Ethereum software that blacklists the address holding the hacker’s stolen funds, making it impossible for him to cash out on the heist.

But this would do nothing to recoup the losses sustained by The DAO. In order to return the funds back to their rightful owners, the Ethereum community would have to agree to a radical maneuver called a “hard fork” which would essentially roll back Ethereum’s historical record of transactions to a point in time before the heist occurred. This option will inevitably inspire heated debate as it calls into question the immutability of the Ethereum blockchain, which is one of the defining characteristics of the technology.

There is no telling yet whether The DAO will survive this hack. But the theft has undeniably harmed the reputation of decentralized financial instruments and the Ethereum project in general. Since Ethereum went live in the spring of 2015, developers have rushed to build autonomously functioning services on the platform. It is now becoming apparent that properly implementing these applications requires expertise in both coding and game theory.

When contracts are not properly vetted, we now know that it is not only the people using that particular application that suffer. The plaform as a whole also takes a hit. In the hours following the heist, the price of Ether has dropped by over 30 percent on online exchanges. 

Remembering When Grand Central Terminal's Trains Went Electric

Despite what the old saying purports, there are new things under the sun. And they get all the press. All too often the old stuff gets short shrift—particularly if it’s under the earth.

Yesterday, at Manhattan’s Grand Central Terminal, attention was paid to the old in an IEEE Milestone ceremony commemorating the electrification of its rail traffic (much of it underground, even then) in 1913. Present at the unveiling of the plaque was John L. Sprague, grandson of Frank J. Sprague, a chief consultant on the project. The elder Sprague was already famous for electrifying the trolleys of Richmond, Va., in the 1880s.

The Milestone was given its due recognition by the IEEE History Center, which is funded in part by the IEEE Foundation.

The electrification project got its start in 1902, when a train engineer, temporarily blinded by smoke from a locomotive’s stack, drove his train into another one, killing 15 people. As a result, the New York State Legislature banned steam-powered locomotives from entering the city after 1908, leaving engineers no alternative to electricity.

In 1903, plans were drawn up, money was appropriated, and train tracks—many already conveniently sited below grade—were encased in tunnels and covered in earth and greenery. The result was what is now known as Park Avenue. The contrast with its smoke-besmirched past jacked up real estate prices—including the associated “air rights,” the sale of which helped to fund the project.

Grand Central was fully operational by 1913, about 11 years after the idea had been proposed. (People worked faster then, didn’t they?) For a stark contrast, look no further than New York City’s Second Avenue subway line, whose first 3-kilometer stretch is supposed to open any day now—about 10 years after workers broke ground.

What’s most striking about the train propulsion technology introduced back then is how little it has changed over the past century. At the plaque-unveiling ceremony, a rusty old segment of the famed, current-carrying third rail was juxtaposed with a new version, and the only visible difference is in the shinier, more conductive alloy. The shape is as Frank Sprague and his colleague William Wilgus described in their patent; the design allows the train’s “shoe” to complete the circuit by touching the rail from underneath. That way, snow and rain can’t easily interfere.

Now is a good time to put up a plaque. Grand Central Terminal is in the midst of a US $210 million facelift, and every bolt is being polished and every slab is being refurbished. Vanderbilt Hall, where the plaque was unveiled, is a marble vault with stately chandeliers and three-story windows worthy of a cathedral. It’s all very old-timey—as is the technology itself.

Not that there’s anything wrong with that. Most technology never dies, and much of it doesn’t even fade away. Instead it continues, right alongside all the newfangled stuff. In his 2006 book, The Shock of the Old: Technology and Global History Since 1900, British historian David Edgerton argues that seemingly futuristic inventions often become museum pieces (think Concorde) while some old standbys just keep on going (think the B-52).  

Nothing has come along to shove electric light rail technology aside, and it’s not for lack of trying. Just ask Elon Musk!

Expert to FBI: Please Join the 21st Century, We Could Use the Help

The FBI needs to update its investigative toolkit and embrace the 21st century, a provocative editorial in this week’s Science magazine argues. The Bureau’s recent squabble with Apple over unlocking a terrorist’s iPhone only underscores the magnitude of the problem, the editorial writer argues. 

In February, the FBI took Apple to court to force the smartphone maker to override the password protection of an iPhone 5C that’d been used by one of the shooters in December’s San Bernadino terrorist attack. Apple held out, saying that caving in would set a dangerous precedent. As both Apple and legal analysts at the time argued, Apple would then have had little recourse but to bow to similar future demands by law enforcement — and repressive governments overseas. 

Read More

Pro-ISIS Online Groups Use Social Media Survival Strategies to Evade Authorities

One of ISIS’ most dangerous weapons has been the Internet. The extremist group relies heavily on social media to spread news and recruit soldiers. And many “lone wolf” criminals including the couple that terrorized San Bernardino, Calif. and the Orlando shooter, are thought to have been radicalized at least in part online. 

Counterterrorism units and law enforcement agencies have struggled to fight back against these new virtual tactics. Now, computer scientists led by Stefan Wuchty at University of Miami have released the first formal survey of how pro-ISIS online groups behave. In particular, their work identifies three popular strategies many groups have used to evade censorship in social media networks. They published their work in Science on Thursday.

Read More

Tech Talk

IEEE Spectrum’s general technology blog, featuring news, analysis, and opinions about engineering, consumer electronics, and technology and society, from the editorial staff and freelance contributors.

Newsletter Sign Up

Sign up for the Tech Alert newsletter and receive ground-breaking technology and science news from IEEE Spectrum every Thursday.

Load More